{
 "generated_at": "2026-10-21T11:54:02Z",
 "rows": [
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 108.21,
   "pp_test": "pp512",
   "tg_tps": 7.92,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-7c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f19d6d1e7f53",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 107.81,
   "pp_test": "pp512",
   "tg_tps": 14.19,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-7c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "24a849a7d72a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 117.25,
   "pp_test": "pp512",
   "tg_tps": 7.91,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ce7722cc426b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 117.96,
   "pp_test": "pp512",
   "tg_tps": 14.15,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a7c648fc12d6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 262.65,
   "pp_test": "pp512",
   "tg_tps": 12.75,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-14c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "18874c039a3d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 235.16,
   "pp_test": "pp512",
   "tg_tps": 21.95,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-14c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "db15c3d57d3a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 232.55,
   "pp_test": "pp512",
   "tg_tps": 35.52,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-14c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ebb460baaab9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 302.14,
   "pp_test": "pp512",
   "tg_tps": 12.75,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4f3c1190754b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 270.37,
   "pp_test": "pp512",
   "tg_tps": 22.34,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "653268ee9d76",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 266.25,
   "pp_test": "pp512",
   "tg_tps": 36.41,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ee045d475e4e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 453.03,
   "pp_test": "pp512",
   "tg_tps": 22.55,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-24c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6a79d0e87ee6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 405.87,
   "pp_test": "pp512",
   "tg_tps": 37.81,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-24c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "80ad55dca886",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 400.26,
   "pp_test": "pp512",
   "tg_tps": 54.61,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-24c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "50a7589115e6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 599.53,
   "pp_test": "pp512",
   "tg_tps": 23.03,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c85e7c2b888d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 537.37,
   "pp_test": "pp512",
   "tg_tps": 40.2,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c20b1ee62b96",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 530.06,
   "pp_test": "pp512",
   "tg_tps": 61.19,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "32ea0952cd91",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 875.81,
   "pp_test": "pp512",
   "tg_tps": 33.92,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-48c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7b3a94ab2707",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 783.45,
   "pp_test": "pp512",
   "tg_tps": 55.69,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-48c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "163b69ce3b1b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 772.24,
   "pp_test": "pp512",
   "tg_tps": 74.93,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-48c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dd59b839126e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1168.89,
   "pp_test": "pp512",
   "tg_tps": 37.01,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-64c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d093ed9fd288",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1042.95,
   "pp_test": "pp512",
   "tg_tps": 59.87,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-64c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "98034ebaf36a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1030.04,
   "pp_test": "pp512",
   "tg_tps": 83.73,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m1-ultra-64c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8dc99211b213",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 147.27,
   "pp_test": "pp512",
   "tg_tps": 12.18,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f5398e2b36e0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 145.91,
   "pp_test": "pp512",
   "tg_tps": 21.7,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "076bba1b8aa9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 201.34,
   "pp_test": "pp512",
   "tg_tps": 6.72,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0166abfdc95e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 181.4,
   "pp_test": "pp512",
   "tg_tps": 12.21,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "38f74ecb907d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 179.57,
   "pp_test": "pp512",
   "tg_tps": 21.91,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f4dadf066268",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 312.65,
   "pp_test": "pp512",
   "tg_tps": 12.47,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "44703fedf188",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 288.46,
   "pp_test": "pp512",
   "tg_tps": 22.7,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a2bc494c6169",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 294.24,
   "pp_test": "pp512",
   "tg_tps": 37.87,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "17f2ea862aa1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 384.38,
   "pp_test": "pp512",
   "tg_tps": 13.06,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-19c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "260069663c9f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 344.5,
   "pp_test": "pp512",
   "tg_tps": 23.01,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-19c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4b8a6843709c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 341.19,
   "pp_test": "pp512",
   "tg_tps": 38.86,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-pro-19c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3c7c1ad300de",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 600.46,
   "pp_test": "pp512",
   "tg_tps": 24.16,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d8c1f27fcab6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 540.15,
   "pp_test": "pp512",
   "tg_tps": 39.97,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4a9075679e29",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 537.6,
   "pp_test": "pp512",
   "tg_tps": 60.99,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "af5a83704103",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 755.67,
   "pp_test": "pp512",
   "tg_tps": 24.65,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-38c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "296163759805",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 677.91,
   "pp_test": "pp512",
   "tg_tps": 41.83,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-38c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ddeb8012da5d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 671.31,
   "pp_test": "pp512",
   "tg_tps": 65.95,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-max-38c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8c9d499e87c2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1128.59,
   "pp_test": "pp512",
   "tg_tps": 39.86,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0a32029a1c4b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1003.16,
   "pp_test": "pp512",
   "tg_tps": 62.14,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3ce7506ca393",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1013.81,
   "pp_test": "pp512",
   "tg_tps": 88.64,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dc32ac976acf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1401.85,
   "pp_test": "pp512",
   "tg_tps": 41.02,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-76c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "03692cf32971",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1248.59,
   "pp_test": "pp512",
   "tg_tps": 66.64,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-76c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a0e47006eb29",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1238.48,
   "pp_test": "pp512",
   "tg_tps": 94.27,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m2-ultra-76c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f3c0a6bb67fb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 187.52,
   "pp_test": "pp512",
   "tg_tps": 12.27,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m3-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "da2f830800f7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 186.75,
   "pp_test": "pp512",
   "tg_tps": 21.34,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m3-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4fccb52e2edc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 272.11,
   "pp_test": "pp512",
   "tg_tps": 17.44,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5aae40d5d871",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 269.49,
   "pp_test": "pp512",
   "tg_tps": 30.65,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4d5247c32cc6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 357.45,
   "pp_test": "pp512",
   "tg_tps": 9.89,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-pro-18c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "06e3325f0b5a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 344.66,
   "pp_test": "pp512",
   "tg_tps": 17.53,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-pro-18c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9da6cc15516d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 341.67,
   "pp_test": "pp512",
   "tg_tps": 30.74,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-pro-18c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "01446c8c5c5c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 589.41,
   "pp_test": "pp512",
   "tg_tps": 19.54,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3f60cc3839ac",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 566.4,
   "pp_test": "pp512",
   "tg_tps": 34.3,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "572a361beb4b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 567.59,
   "pp_test": "pp512",
   "tg_tps": 56.58,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-30c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "19b00ad09305",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 779.17,
   "pp_test": "pp512",
   "tg_tps": 25.09,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c42058133237",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 757.64,
   "pp_test": "pp512",
   "tg_tps": 42.75,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1c9d7e356bab",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 759.7,
   "pp_test": "pp512",
   "tg_tps": 66.31,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e72c4befc9f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1121.8,
   "pp_test": "pp512",
   "tg_tps": 42.24,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "daa0a7451b3f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1085.76,
   "pp_test": "pp512",
   "tg_tps": 63.55,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b629417167c5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1073.09,
   "pp_test": "pp512",
   "tg_tps": 88.4,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-60c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "22c9adce4036",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1538.34,
   "pp_test": "pp512",
   "tg_tps": 39.78,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2eede9058895",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1487.51,
   "pp_test": "pp512",
   "tg_tps": 63.93,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bf2bc60f17d5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1471.24,
   "pp_test": "pp512",
   "tg_tps": 92.14,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7a168685d4bb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 230.18,
   "pp_test": "pp512",
   "tg_tps": 7.43,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c5b11db3dd53",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 223.64,
   "pp_test": "pp512",
   "tg_tps": 13.54,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "da84cd59864b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 221.29,
   "pp_test": "pp512",
   "tg_tps": 24.11,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "722b5955e28d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 381.14,
   "pp_test": "pp512",
   "tg_tps": 17.19,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "010b0caf5e03",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 367.13,
   "pp_test": "pp512",
   "tg_tps": 30.54,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "35ea2888f49d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 364.06,
   "pp_test": "pp512",
   "tg_tps": 49.64,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "80c0c6665627",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 464.48,
   "pp_test": "pp512",
   "tg_tps": 17.18,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e9ef9ec26d96",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 449.62,
   "pp_test": "pp512",
   "tg_tps": 30.69,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4c29d933c7eb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 439.78,
   "pp_test": "pp512",
   "tg_tps": 50.74,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b9bc67af2e0a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 736.25,
   "pp_test": "pp512",
   "tg_tps": 24.29,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b8ff0ed405de",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 718.56,
   "pp_test": "pp512",
   "tg_tps": 43.87,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0796d15548ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 713.93,
   "pp_test": "pp512",
   "tg_tps": 69.95,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-32c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "53ba062d78a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 922.83,
   "pp_test": "pp512",
   "tg_tps": 31.64,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "24dadc53d617",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 891.94,
   "pp_test": "pp512",
   "tg_tps": 54.05,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0cce3e2ab40e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 885.68,
   "pp_test": "pp512",
   "tg_tps": 83.06,
   "tg_test": "tg128",
   "commit": "8e672ef",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "768b3c157de8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 715.42,
   "pp_test": "pp512",
   "tg_tps": 18.42,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m5-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "326ea6463e1e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 722.79,
   "pp_test": "pp512",
   "tg_tps": 31.88,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": "partial row in source",
   "hardware": "apple-m5-10c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a2e4026ef3d9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1310.78,
   "pp_test": "pp512",
   "tg_tps": 21.39,
   "tg_test": "tg128",
   "commit": "60081bb",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1d4c3190af5f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1302.59,
   "pp_test": "pp512",
   "tg_tps": 38.94,
   "tg_test": "tg128",
   "commit": "60081bb",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e37e3313d8a8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1340.44,
   "pp_test": "pp512",
   "tg_tps": 67.59,
   "tg_test": "tg128",
   "commit": "60081bb",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-16c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c13a2b1a1838",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1588.78,
   "pp_test": "pp512",
   "tg_tps": 21.55,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4a1398aae6f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1553.86,
   "pp_test": "pp512",
   "tg_tps": 38.92,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7d462b0f8fa4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Pro",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 1620.64,
   "pp_test": "pp512",
   "tg_tps": 66.33,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-pro-20c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "05f41c5f04b4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Max",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 3158.49,
   "pp_test": "pp512",
   "tg_tps": 37.11,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d890ad47f816",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Max",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 3143.81,
   "pp_test": "pp512",
   "tg_tps": 72.42,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "999f07642bbe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Max",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 3219.99,
   "pp_test": "pp512",
   "tg_tps": 119.92,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-max-40c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4d27f090d751",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Ultra",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 5207.14,
   "pp_test": "pp512",
   "tg_tps": 73.6,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "30a9efbf9cc0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 4910.55,
   "pp_test": "pp512",
   "tg_tps": 119.01,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7c8dd320287f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M5 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 4944.66,
   "pp_test": "pp512",
   "tg_tps": 179.1,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m5-ultra-80c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "026603ee65f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M6",
   "quant": "F16",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 952.53,
   "pp_test": "pp512",
   "tg_tps": 11.59,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m6-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ef243311c5b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M6",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 970.46,
   "pp_test": "pp512",
   "tg_tps": 21.01,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m6-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e9a98dc615db",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-4167",
   "source_name": "llama.cpp Apple Silicon thread (#4167)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/4167",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M6",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": false,
   "pp_tps": 988.93,
   "pp_test": "pp512",
   "tg_tps": 35.74,
   "tg_test": "tg128",
   "commit": "c1d0e7a",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m6-8c",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c9fb9717330e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 14073.41,
   "pp_test": "pp512",
   "tg_tps": 290.02,
   "tg_test": "tg128",
   "commit": "8cf6b42",
   "contributor": "@totaldev",
   "notes": null,
   "hardware": "nvidia-rtx-5090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f2d3d9e5cc83",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX PRO 6000 Blackwell",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 14854.63,
   "pp_test": "pp512",
   "tg_tps": 274.2,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@Tom94",
   "notes": null,
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dccb809e928c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "H100 80 GB",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 9918.34,
   "pp_test": "pp512",
   "tg_tps": 267.81,
   "tg_test": "tg128",
   "commit": "5143fa8",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-h100-sxm",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "024df9c1e5f1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A100 80 GB",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 4849.53,
   "pp_test": "pp512",
   "tg_tps": 190.88,
   "tg_test": "tg128",
   "commit": "5143fa8",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a100-80gb-pcie",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "03558f1aa9ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4090 D",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 10293.86,
   "pp_test": "pp512",
   "tg_tps": 189.33,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@autonomous-ai-lab",
   "notes": null,
   "hardware": "nvidia-rtx-4090-d",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6d5d3ef33e4c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 11992.7,
   "pp_test": "pp512",
   "tg_tps": 186.21,
   "tg_test": "tg128",
   "commit": "2241453",
   "contributor": "@lhl",
   "notes": null,
   "hardware": "nvidia-rtx-4090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1b3830cfa90a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 8297.36,
   "pp_test": "pp512",
   "tg_tps": 181.99,
   "tg_test": "tg128",
   "commit": "8a4280c",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-5080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fce82a3037c8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5070 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 6952.38,
   "pp_test": "pp512",
   "tg_tps": 176.85,
   "tg_test": "tg128",
   "commit": "933414c",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "83ee2ee3f594",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 6000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 9229.23,
   "pp_test": "pp512",
   "tg_tps": 176.07,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "546ff0f3462c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3090 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 6567.49,
   "pp_test": "pp512",
   "tg_tps": 171.19,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@slaren",
   "notes": null,
   "hardware": "nvidia-rtx-3090-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b722cf724577",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 5174.69,
   "pp_test": "pp512",
   "tg_tps": 158.16,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-rtx-3090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "141a96423e03",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5090 Laptop",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 6667.06,
   "pp_test": "pp512",
   "tg_tps": 156.49,
   "tg_test": "tg128",
   "commit": "29f1482",
   "contributor": "@rs38",
   "notes": null,
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "71654ce36f10",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "L40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 8870.49,
   "pp_test": "pp512",
   "tg_tps": 152.01,
   "tg_test": "tg128",
   "commit": "ee09828",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-l40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d29aa62c4127",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4080 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 8125.15,
   "pp_test": "pp512",
   "tg_tps": 148.33,
   "tg_test": "tg128",
   "commit": "81086cd",
   "contributor": "@zacharyarnaise",
   "notes": null,
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f98362649c3e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 8031.64,
   "pp_test": "pp512",
   "tg_tps": 142.49,
   "tg_test": "tg128",
   "commit": "20638e4",
   "contributor": "@Ristovski",
   "notes": null,
   "hardware": "nvidia-rtx-4080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "15d8262df1ba",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 5013.86,
   "pp_test": "pp512",
   "tg_tps": 139.65,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@slaren",
   "notes": null,
   "hardware": "nvidia-rtx-3080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cd278ba78226",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A6000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 4913.93,
   "pp_test": "pp512",
   "tg_tps": 138.73,
   "tg_test": "tg128",
   "commit": "4795c91",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-a6000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "51c09ac21bdd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4070 Ti SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 6924.53,
   "pp_test": "pp512",
   "tg_tps": 132.26,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@Ristovski",
   "notes": null,
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "855548eb1555",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX PRO 4000 Blackwell",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 4992.83,
   "pp_test": "pp512",
   "tg_tps": 131.66,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-pro-4000-blackwell",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "54f3cc7a847e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A5000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 4028.16,
   "pp_test": "pp512",
   "tg_tps": 130.07,
   "tg_test": "tg128",
   "commit": "e5155e6",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "eb10fe32a93d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla V100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 3042.64,
   "pp_test": "pp512",
   "tg_tps": 129.08,
   "tg_test": "tg128",
   "commit": "51f5a45",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "afbd4b786c4c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5070",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 5184.75,
   "pp_test": "pp512",
   "tg_tps": 127.54,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "@Spyro000",
   "notes": null,
   "hardware": "nvidia-rtx-5070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4b0dd6c390d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 4609.01,
   "pp_test": "pp512",
   "tg_tps": 124.11,
   "tg_test": "tg128",
   "commit": "3470a5c",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4423001a7d53",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A30",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2767.1,
   "pp_test": "pp512",
   "tg_tps": 124.81,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a30",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7fa28a978924",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Titan V",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2617.46,
   "pp_test": "pp512",
   "tg_tps": 108.79,
   "tg_test": "tg128",
   "commit": "e56abd2",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-titan-v",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "336484901268",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2080 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2890.66,
   "pp_test": "pp512",
   "tg_tps": 107.51,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "055bcd69f0ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 6000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2751.18,
   "pp_test": "pp512",
   "tg_tps": 102.77,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-6000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "61b181e24e9f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2709.95,
   "pp_test": "pp512",
   "tg_tps": 102.68,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9c7e1a9c0243",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4500",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2827.2,
   "pp_test": "pp512",
   "tg_tps": 97.32,
   "tg_test": "tg128",
   "commit": "5cdb27e",
   "contributor": "@aleksyx",
   "notes": null,
   "hardware": "nvidia-rtx-a4500",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "15a2492102e5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5060 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 3737.25,
   "pp_test": "pp512",
   "tg_tps": 90.94,
   "tg_test": "tg128",
   "commit": "89d1029",
   "contributor": "@mike-llamacpp",
   "notes": null,
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ab1c11daf292",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2070 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2088.34,
   "pp_test": "pp512",
   "tg_tps": 88.06,
   "tg_test": "tg128",
   "commit": "bc07349",
   "contributor": "@phstudy",
   "notes": null,
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7a7f1257043a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2684.06,
   "pp_test": "pp512",
   "tg_tps": 83.77,
   "tg_test": "tg128",
   "commit": "65349f2",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "10a37abe41c7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Titan Xp",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1154.96,
   "pp_test": "pp512",
   "tg_tps": 76.08,
   "tg_test": "tg128",
   "commit": "c4510dc",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-titan-xp",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "825e16a47239",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3060",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2137.5,
   "pp_test": "pp512",
   "tg_tps": 75.57,
   "tg_test": "tg128",
   "commit": "baa9255",
   "contributor": "@QuantiusBenignus",
   "notes": null,
   "hardware": "nvidia-rtx-3060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "53a016f36808",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 4000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1536.89,
   "pp_test": "pp512",
   "tg_tps": 65.62,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f18eea3acf0f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4060 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 3394.63,
   "pp_test": "pp512",
   "tg_tps": 63.86,
   "tg_test": "tg128",
   "commit": "89d1029",
   "contributor": "@mike-llamacpp",
   "notes": null,
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ca389b8f6d41",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1080 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1084.41,
   "pp_test": "pp512",
   "tg_tps": 62.49,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-gtx-1080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "408240a8a71c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 2779.77,
   "pp_test": "pp512",
   "tg_tps": 61.83,
   "tg_test": "tg128",
   "commit": "a74a0d6",
   "contributor": "@sdwolfz",
   "notes": null,
   "hardware": "nvidia-rtx-4000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "863ac6c7b46c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2060 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1420.24,
   "pp_test": "pp512",
   "tg_tps": 60.04,
   "tg_test": "tg128",
   "commit": "5c0eb5e",
   "contributor": "@ggerganov",
   "notes": null,
   "hardware": "nvidia-rtx-2060-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0e2dd34b6a57",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 760.8,
   "pp_test": "pp512",
   "tg_tps": 58.35,
   "tg_test": "tg128",
   "commit": "b8372ee",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-p100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c0b4d123cac4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "DGX Spark",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 3062.31,
   "pp_test": "pp512",
   "tg_tps": 57.21,
   "tg_test": "tg128",
   "commit": "5acd455",
   "contributor": "@ggerganov",
   "notes": null,
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "80302e1f32d3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1007.42,
   "pp_test": "pp512",
   "tg_tps": 54.74,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-tesla-p40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2b504bef72b7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1956.22,
   "pp_test": "pp512",
   "tg_tps": 50.62,
   "tg_test": "tg128",
   "commit": "756cfea",
   "contributor": "@DigitalRudeness",
   "notes": null,
   "hardware": "nvidia-rtx-2000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ea096d2bdb00",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla T4",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1219.06,
   "pp_test": "pp512",
   "tg_tps": 46.38,
   "tg_test": "tg128",
   "commit": "d32e03f",
   "contributor": "@pt13762104",
   "notes": null,
   "hardware": "nvidia-tesla-t4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "948c87fcfdfc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4050 Laptop",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 1725.85,
   "pp_test": "pp512",
   "tg_tps": 43.72,
   "tg_test": "tg128",
   "commit": "d79d8f3",
   "contributor": "@TimCabbage",
   "notes": null,
   "hardware": "nvidia-rtx-4050-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "02681baec184",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1660",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 148.91,
   "pp_test": "pp512",
   "tg_tps": 41.35,
   "tg_test": "tg128",
   "commit": "9515c61",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-gtx-1660",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ccf61f689896",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla M40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 282.65,
   "pp_test": "pp512",
   "tg_tps": 38.04,
   "tg_test": "tg128",
   "commit": "97d5117",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-m40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "da12e060528b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1070 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 714.44,
   "pp_test": "pp512",
   "tg_tps": 37.82,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-gtx-1070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f48b7904c4cd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Jetson AGX Orin",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 991.31,
   "pp_test": "pp512",
   "tg_tps": 33.58,
   "tg_test": "tg128",
   "commit": "c1b1876",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-jetson-agx-orin-64gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "028bf0eeb916",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P4",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 514.53,
   "pp_test": "pp512",
   "tg_tps": 33.29,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-tesla-p4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "038a484ac8d6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "P106-100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 406.94,
   "pp_test": "pp512",
   "tg_tps": 30.4,
   "tg_test": "tg128",
   "commit": "5fd160b",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-p106-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "989f90df0992",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1060",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 416.85,
   "pp_test": "pp512",
   "tg_tps": 27.79,
   "tg_test": "tg128",
   "commit": "5fd160b",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-gtx-1060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "14f85c13119d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro T1000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 79.44,
   "pp_test": "pp512",
   "tg_tps": 27.82,
   "tg_test": "tg128",
   "commit": "f6da8cb",
   "contributor": "@hanabu",
   "notes": null,
   "hardware": "nvidia-t1000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dd33dd1c581a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro P2000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 309.3,
   "pp_test": "pp512",
   "tg_tps": 23.63,
   "tg_test": "tg128",
   "commit": "baa9255",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-quadro-p2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c388de616285",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro P1000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 183.4,
   "pp_test": "pp512",
   "tg_tps": 13.99,
   "tg_test": "tg128",
   "commit": "1e74897",
   "contributor": "@aleksyx",
   "notes": null,
   "hardware": "nvidia-quadro-p1000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "49e3d1e050ea",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla K80",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": false,
   "pp_tps": 133.14,
   "pp_test": "pp512",
   "tg_tps": 13.8,
   "tg_test": "tg128",
   "commit": "32732f2",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-tesla-k80",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0233660ec8e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 14970.15,
   "pp_test": "pp512",
   "tg_tps": 300.4,
   "tg_test": "tg128",
   "commit": "8cf6b42",
   "contributor": "@totaldev",
   "notes": null,
   "hardware": "nvidia-rtx-5090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "23833a7454f0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX PRO 6000 Blackwell",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 16618.98,
   "pp_test": "pp512",
   "tg_tps": 281.11,
   "tg_test": "tg128",
   "commit": "5143fa8",
   "contributor": "@Tom94",
   "notes": null,
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0aa3076b13fd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "H100 80 GB",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 11263.29,
   "pp_test": "pp512",
   "tg_tps": 280.74,
   "tg_test": "tg128",
   "commit": "5143fa8",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-h100-sxm",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c2bd95d25351",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A100 80 GB",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5285.96,
   "pp_test": "pp512",
   "tg_tps": 200.9,
   "tg_test": "tg128",
   "commit": "5143fa8",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a100-80gb-pcie",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "93a973218648",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4090 D",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 12506.97,
   "pp_test": "pp512",
   "tg_tps": 191.57,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@autonomous-ai-lab",
   "notes": null,
   "hardware": "nvidia-rtx-4090-d",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "442236158e78",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 14770.63,
   "pp_test": "pp512",
   "tg_tps": 188.96,
   "tg_test": "tg128",
   "commit": "2241453",
   "contributor": "@lhl",
   "notes": null,
   "hardware": "nvidia-rtx-4090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4b44496bee6e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9487.7,
   "pp_test": "pp512",
   "tg_tps": 184.68,
   "tg_test": "tg128",
   "commit": "8a4280c",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-5080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5bc2982c8f74",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5070 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 8419.56,
   "pp_test": "pp512",
   "tg_tps": 182.43,
   "tg_test": "tg128",
   "commit": "933414c",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e0a9f00fe6c1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 6000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 10576.85,
   "pp_test": "pp512",
   "tg_tps": 179.47,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "71de6779c23b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3090 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 6924.01,
   "pp_test": "pp512",
   "tg_tps": 172.26,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@slaren",
   "notes": null,
   "hardware": "nvidia-rtx-3090-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "620493c712d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX PRO 4500 Blackwell",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 7251.66,
   "pp_test": "pp512",
   "tg_tps": 168.9,
   "tg_test": "tg128",
   "commit": "becc481",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-pro-4500-blackwell",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c960fe866886",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3090",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5560.06,
   "pp_test": "pp512",
   "tg_tps": 161.89,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-rtx-3090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "69ae521caeb9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5090 Laptop",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 7641.89,
   "pp_test": "pp512",
   "tg_tps": 158.14,
   "tg_test": "tg128",
   "commit": "29f1482",
   "contributor": "@rs38",
   "notes": null,
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8afa605ff6e1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "L40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 10097.64,
   "pp_test": "pp512",
   "tg_tps": 153.76,
   "tg_test": "tg128",
   "commit": "ee09828",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-l40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "de5a539cbfd6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4080 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9439.01,
   "pp_test": "pp512",
   "tg_tps": 147.48,
   "tg_test": "tg128",
   "commit": "81086cd",
   "contributor": "@zacharyarnaise",
   "notes": null,
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a2a42fcbb24b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9205.93,
   "pp_test": "pp512",
   "tg_tps": 143.47,
   "tg_test": "tg128",
   "commit": "20638e4",
   "contributor": "@Ristovski",
   "notes": null,
   "hardware": "nvidia-rtx-4080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a52cf8860f08",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A6000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5662.39,
   "pp_test": "pp512",
   "tg_tps": 144.87,
   "tg_test": "tg128",
   "commit": "4795c91",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-a6000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9e3ccec1a639",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3080",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5569.56,
   "pp_test": "pp512",
   "tg_tps": 139.95,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@slaren",
   "notes": null,
   "hardware": "nvidia-rtx-3080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7a4f6e639b4d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX PRO 4000 Blackwell",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5674.44,
   "pp_test": "pp512",
   "tg_tps": 136.38,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-pro-4000-blackwell",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f51954236a6d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A5000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 4552.15,
   "pp_test": "pp512",
   "tg_tps": 135.83,
   "tg_test": "tg128",
   "commit": "e5155e6",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3dac9366f375",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla V100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2973.78,
   "pp_test": "pp512",
   "tg_tps": 134.76,
   "tg_test": "tg128",
   "commit": "51f5a45",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "53f1d34c99a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4070 Ti SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 7612.32,
   "pp_test": "pp512",
   "tg_tps": 132.85,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@Ristovski",
   "notes": null,
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5ea52b6e8943",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5070",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5783.44,
   "pp_test": "pp512",
   "tg_tps": 128.21,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "@Spyro000",
   "notes": null,
   "hardware": "nvidia-rtx-5070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f3c1cceb4098",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5256.38,
   "pp_test": "pp512",
   "tg_tps": 126.24,
   "tg_test": "tg128",
   "commit": "3470a5c",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "98dd1a3f346e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "A30",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3068.72,
   "pp_test": "pp512",
   "tg_tps": 131.93,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-a30",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1c0d176b6dbd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Titan V",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2481.25,
   "pp_test": "pp512",
   "tg_tps": 112.17,
   "tg_test": "tg128",
   "commit": "e56abd2",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-titan-v",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "893350082daa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2080 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3107.61,
   "pp_test": "pp512",
   "tg_tps": 109.17,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4862be6138e0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 6000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3053.96,
   "pp_test": "pp512",
   "tg_tps": 104.38,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-6000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "20e748757f45",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3052.35,
   "pp_test": "pp512",
   "tg_tps": 103.63,
   "tg_test": "tg128",
   "commit": "b8e09f0",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cb7ea6dc2496",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4500",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3453.1,
   "pp_test": "pp512",
   "tg_tps": 103.0,
   "tg_test": "tg128",
   "commit": "5cdb27e",
   "contributor": "@aleksyx",
   "notes": null,
   "hardware": "nvidia-rtx-a4500",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "122813425d77",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 5060 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 4195.53,
   "pp_test": "pp512",
   "tg_tps": 93.46,
   "tg_test": "tg128",
   "commit": "89d1029",
   "contributor": "@mike-llamacpp",
   "notes": null,
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "30c040a7c646",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2070 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2293.29,
   "pp_test": "pp512",
   "tg_tps": 87.71,
   "tg_test": "tg128",
   "commit": "bc07349",
   "contributor": "@phstudy",
   "notes": null,
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7e9b59e73a2f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2807.83,
   "pp_test": "pp512",
   "tg_tps": 85.17,
   "tg_test": "tg128",
   "commit": "65349f2",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e9adb08f1bbd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 3060",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2407.67,
   "pp_test": "pp512",
   "tg_tps": 76.92,
   "tg_test": "tg128",
   "commit": "baa9255",
   "contributor": "@QuantiusBenignus",
   "notes": null,
   "hardware": "nvidia-rtx-3060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "55ade40f4357",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Titan Xp",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1218.12,
   "pp_test": "pp512",
   "tg_tps": 73.84,
   "tg_test": "tg128",
   "commit": "c4510dc",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-titan-xp",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c1deda0f5677",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro RTX 4000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1662.8,
   "pp_test": "pp512",
   "tg_tps": 67.62,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-quadro-rtx-4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "020e250fed1c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 4060 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3803.45,
   "pp_test": "pp512",
   "tg_tps": 64.03,
   "tg_test": "tg128",
   "commit": "89d1029",
   "contributor": "@mike-llamacpp",
   "notes": null,
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "878638fee23f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX A4000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3171.86,
   "pp_test": "pp512",
   "tg_tps": 61.37,
   "tg_test": "tg128",
   "commit": "a74a0d6",
   "contributor": "@sdwolfz",
   "notes": null,
   "hardware": "nvidia-rtx-4000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b1f7de184b84",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 787.36,
   "pp_test": "pp512",
   "tg_tps": 61.99,
   "tg_test": "tg128",
   "commit": "b8372ee",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-p100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2976e010ac6d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1080 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1138.14,
   "pp_test": "pp512",
   "tg_tps": 61.38,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-gtx-1080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7c86524aa3cc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2060 SUPER",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1563.77,
   "pp_test": "pp512",
   "tg_tps": 61.13,
   "tg_test": "tg128",
   "commit": "5c0eb5e",
   "contributor": "@ggerganov",
   "notes": null,
   "hardware": "nvidia-rtx-2060-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "302cc15413fa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "DGX Spark",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3661.37,
   "pp_test": "pp512",
   "tg_tps": 56.74,
   "tg_test": "tg128",
   "commit": "5acd455",
   "contributor": "@ggerganov",
   "notes": null,
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "84294c6202e3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1079.66,
   "pp_test": "pp512",
   "tg_tps": 53.73,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-tesla-p40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0f0f0da1cfa3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RTX 2000 Ada",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2250.14,
   "pp_test": "pp512",
   "tg_tps": 50.71,
   "tg_test": "tg128",
   "commit": "756cfea",
   "contributor": "@DigitalRudeness",
   "notes": null,
   "hardware": "nvidia-rtx-2000-ada",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a2e000e5f710",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla T4",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1309.73,
   "pp_test": "pp512",
   "tg_tps": 44.03,
   "tg_test": "tg128",
   "commit": "d32e03f",
   "contributor": "@pt13762104",
   "notes": null,
   "hardware": "nvidia-tesla-t4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c78382e6a944",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1660",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 154.45,
   "pp_test": "pp512",
   "tg_tps": 41.43,
   "tg_test": "tg128",
   "commit": "9515c61",
   "contributor": "@ariya",
   "notes": null,
   "hardware": "nvidia-gtx-1660",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b809a814430f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla M40",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 290.17,
   "pp_test": "pp512",
   "tg_tps": 39.98,
   "tg_test": "tg128",
   "commit": "97d5117",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "nvidia-tesla-m40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "565f4249ced9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1070 Ti",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 790.52,
   "pp_test": "pp512",
   "tg_tps": 37.87,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-gtx-1070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ba803695c2b4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Jetson AGX Orin",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1171.96,
   "pp_test": "pp512",
   "tg_tps": 35.88,
   "tg_test": "tg128",
   "commit": "c1b1876",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-jetson-agx-orin-64gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b7fe2f52207b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla P4",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 529.53,
   "pp_test": "pp512",
   "tg_tps": 33.12,
   "tg_test": "tg128",
   "commit": "c76b420",
   "contributor": "@m18coppola",
   "notes": null,
   "hardware": "nvidia-tesla-p4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4e820ac7af33",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "P106-100",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 438.49,
   "pp_test": "pp512",
   "tg_tps": 30.64,
   "tg_test": "tg128",
   "commit": "5fd160b",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-p106-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3e37f07c13d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "GTX 1060",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 446.19,
   "pp_test": "pp512",
   "tg_tps": 28.18,
   "tg_test": "tg128",
   "commit": "5fd160b",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-gtx-1060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8c8e7542b48f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro T1000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 27.46,
   "pp_test": "pp512",
   "tg_tps": 27.46,
   "tg_test": "tg128",
   "commit": "f6da8cb",
   "contributor": "@hanabu",
   "notes": null,
   "hardware": "nvidia-t1000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "07328d73b866",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro P2000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 311.55,
   "pp_test": "pp512",
   "tg_tps": 23.76,
   "tg_test": "tg128",
   "commit": "baa9255",
   "contributor": "@TinyServal",
   "notes": null,
   "hardware": "nvidia-quadro-p2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4f0ecac6a0f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Tesla K80",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 133.36,
   "pp_test": "pp512",
   "tg_tps": 14.27,
   "tg_test": "tg128",
   "commit": "32732f2",
   "contributor": "@pebaryan",
   "notes": null,
   "hardware": "nvidia-tesla-k80",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "69645600889f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15013",
   "source_name": "llama.cpp CUDA scoreboard (#15013)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15013",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Quadro P1000",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 173.82,
   "pp_test": "pp512",
   "tg_tps": 13.65,
   "tg_test": "tg128",
   "commit": "1e74897",
   "contributor": "@aleksyx",
   "notes": null,
   "hardware": "nvidia-quadro-p1000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1adc3621b5d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 10381.64,
   "pp_test": "pp512",
   "tg_tps": 263.63,
   "tg_test": "tg128",
   "commit": "ca71fb9",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c5d86e49d91f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 9452.03,
   "pp_test": "pp512",
   "tg_tps": 187.97,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7a0ec8a3fac3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5080",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 7444.99,
   "pp_test": "pp512",
   "tg_tps": 185.1,
   "tg_test": "tg128",
   "commit": "f6b533d",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fbc6dff1e197",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 XTX",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3726.99,
   "pp_test": "pp512",
   "tg_tps": 182.63,
   "tg_test": "tg128",
   "commit": "304665f",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7b99f4c43e90",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 4666.15,
   "pp_test": "pp512",
   "tg_tps": 164.05,
   "tg_test": "tg128",
   "commit": "d05fe1d",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "71f4fc6eebb3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia A100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 6389.86,
   "pp_test": "pp512",
   "tg_tps": 160.78,
   "tg_test": "tg128",
   "commit": "2257758",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-a100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6d635f07a651",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5090 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 5546.39,
   "pp_test": "pp512",
   "tg_tps": 149.41,
   "tg_test": "tg128",
   "commit": "4bc272f",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8e108cbb8326",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4080 Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 7101.18,
   "pp_test": "pp512",
   "tg_tps": 147.13,
   "tg_test": "tg128",
   "commit": "81086cd",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "96487a6a7106",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon AI Pro R9700",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 5609.82,
   "pp_test": "pp512",
   "tg_tps": 145.67,
   "tg_test": "tg128",
   "commit": "dd1ea52",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-ai-pro-r9700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7ee76c5b2f48",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3080",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 4287.11,
   "pp_test": "pp512",
   "tg_tps": 139.15,
   "tg_test": "tg128",
   "commit": "7c7d6ce",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "25bb180468a3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A5000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3641.55,
   "pp_test": "pp512",
   "tg_tps": 139.89,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a6419f561fa4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9070 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 5036.04,
   "pp_test": "pp512",
   "tg_tps": 137.11,
   "tg_test": "tg128",
   "commit": "e9fd8dc",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9070-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c0b5990cac84",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 6213.63,
   "pp_test": "pp512",
   "tg_tps": 135.63,
   "tg_test": "tg128",
   "commit": "d13d0f6",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "89a0d517a4d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla V100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1391.39,
   "pp_test": "pp512",
   "tg_tps": 129.58,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "79d68fde6c63",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Ti Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 6099.18,
   "pp_test": "pp512",
   "tg_tps": 129.45,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f397675392f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2941.58,
   "pp_test": "pp512",
   "tg_tps": 123.18,
   "tg_test": "tg128",
   "commit": "71e74a3",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "35d7fedd4caf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3164.1,
   "pp_test": "pp512",
   "tg_tps": 119.71,
   "tg_test": "tg128",
   "commit": "21c17b5",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "30a17f88734b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7800 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2017.33,
   "pp_test": "pp512",
   "tg_tps": 118.27,
   "tg_test": "tg128",
   "commit": "4fdbc1e",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "49fbc503ee31",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 GRE",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2336.31,
   "pp_test": "pp512",
   "tg_tps": 116.11,
   "tg_test": "tg128",
   "commit": "4b2a477",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-gre",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ade70d41f01f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3 Ultra",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1116.83,
   "pp_test": "pp512",
   "tg_tps": 115.54,
   "tg_test": "tg128",
   "commit": "2d451c8",
   "contributor": null,
   "notes": "MoltenVK",
   "hardware": "apple-m3-ultra",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9e2588a0597b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B70",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3379.0,
   "pp_test": "pp512",
   "tg_tps": 112.02,
   "tg_test": "tg128",
   "commit": "b863507",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b70",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "16025f3b7815",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia A30",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3290.7,
   "pp_test": "pp512",
   "tg_tps": 111.48,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-a30",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "713b6c097895",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 4981.44,
   "pp_test": "pp512",
   "tg_tps": 110.53,
   "tg_test": "tg128",
   "commit": "516a4ca",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b3aa0ce6fe7a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 4608.2,
   "pp_test": "pp512",
   "tg_tps": 108.74,
   "tg_test": "tg128",
   "commit": "c945aaa",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f36c95b19c22",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Instinct MI50",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1119.55,
   "pp_test": "pp512",
   "tg_tps": 108.51,
   "tg_test": "tg128",
   "commit": "3af34b9",
   "contributor": null,
   "notes": null,
   "hardware": "amd-instinct-mi50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "af05200d6a1b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6900 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1901.2,
   "pp_test": "pp512",
   "tg_tps": 108.0,
   "tg_test": "tg128",
   "commit": "a972fae",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1d1037f78f4e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro VII",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 912.47,
   "pp_test": "pp512",
   "tg_tps": 106.03,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-vii",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5540c089676e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Titan V",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 796.29,
   "pp_test": "pp512",
   "tg_tps": 105.06,
   "tg_test": "tg128",
   "commit": "e56abd2",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-titan-v",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f7f7189bd033",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1593.97,
   "pp_test": "pp512",
   "tg_tps": 101.47,
   "tg_test": "tg128",
   "commit": "f1cee99",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5056a09d424b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon VII",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1059.14,
   "pp_test": "pp512",
   "tg_tps": 101.19,
   "tg_test": "tg128",
   "commit": "77d6ae4",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-vii",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d0b3654dba4b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1752.92,
   "pp_test": "pp512",
   "tg_tps": 100.32,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "123902afae51",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 2080 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1888.24,
   "pp_test": "pp512",
   "tg_tps": 97.58,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "beec81cebb24",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W6800X Duo",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 687.71,
   "pp_test": "pp512",
   "tg_tps": 94.82,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-w6800x-duo",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6049d5f2c18c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5060 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3460.92,
   "pp_test": "pp512",
   "tg_tps": 93.51,
   "tg_test": "tg128",
   "commit": "89f10ba",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "86c5fead4ee7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 3179.37,
   "pp_test": "pp512",
   "tg_tps": 92.29,
   "tg_test": "tg128",
   "commit": "9a48399",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-4070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9cbec9ad6b37",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3060 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2361.43,
   "pp_test": "pp512",
   "tg_tps": 86.74,
   "tg_test": "tg128",
   "commit": "b42c7fa",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3dff43fbf114",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W6800X",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 510.8,
   "pp_test": "pp512",
   "tg_tps": 86.47,
   "tg_test": "tg128",
   "commit": "13b4548",
   "contributor": null,
   "notes": "MoltenVK",
   "hardware": "amd-radeon-pro-w6800x",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4e2a94fbdc2b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6700 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1051.2,
   "pp_test": "pp512",
   "tg_tps": 83.88,
   "tg_test": "tg128",
   "commit": "6d75883",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "86cae0c9ace4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6750 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1040.58,
   "pp_test": "pp512",
   "tg_tps": 81.98,
   "tg_test": "tg128",
   "commit": "228f34c",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6750-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bc09a690b888",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro V620",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1595.32,
   "pp_test": "pp512",
   "tg_tps": 81.78,
   "tg_test": "tg128",
   "commit": "03d4698",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8f44a2949564",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2113.02,
   "pp_test": "pp512",
   "tg_tps": 78.71,
   "tg_test": "tg128",
   "commit": "1b8fb81",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-3070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fec7f84d8ecc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Instinct MI60",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 369.26,
   "pp_test": "pp512",
   "tg_tps": 78.16,
   "tg_test": "tg128",
   "commit": "504af20",
   "contributor": null,
   "notes": null,
   "hardware": "amd-instinct-mi60",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "42b11e483002",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3060",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1815.7,
   "pp_test": "pp512",
   "tg_tps": 75.94,
   "tg_test": "tg128",
   "commit": "92c0b38",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d49bfb7523c2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 724.77,
   "pp_test": "pp512",
   "tg_tps": 75.02,
   "tg_test": "tg128",
   "commit": "1ece0cb6",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "66f4c1fda321",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla T10",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1692.7,
   "pp_test": "pp512",
   "tg_tps": 75.01,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-tesla-t10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c727f0f0220d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A4000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2248.14,
   "pp_test": "pp512",
   "tg_tps": 73.74,
   "tg_test": "tg128",
   "commit": "f5245b5",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b21faadeb8f2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9060 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2844.8,
   "pp_test": "pp512",
   "tg_tps": 71.07,
   "tg_test": "tg128",
   "commit": "f1cee99",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9060-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "04b9f7697687",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 5700 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 529.69,
   "pp_test": "pp512",
   "tg_tps": 70.73,
   "tg_test": "tg128",
   "commit": "4fdbc1e",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-5700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "652bf10bee8d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc B580",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 620.94,
   "pp_test": "pp512",
   "tg_tps": 70.14,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-b580",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "912cd910e18a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro V540",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 583.88,
   "pp_test": "pp512",
   "tg_tps": 69.64,
   "tg_test": "tg128",
   "commit": "9da3dcd",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-v540",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0f1444173a51",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W5700",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 449.85,
   "pp_test": "pp512",
   "tg_tps": 68.55,
   "tg_test": "tg128",
   "commit": "23bc779",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-w5700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a0f9132962f6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B60",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 522.36,
   "pp_test": "pp512",
   "tg_tps": 68.55,
   "tg_test": "tg128",
   "commit": "516a4ca",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b60",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "38e39ce14a7f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1080 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 585.54,
   "pp_test": "pp512",
   "tg_tps": 67.75,
   "tg_test": "tg128",
   "commit": "ac2557c",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b9d3c7ed6295",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 2070 Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1199.13,
   "pp_test": "pp512",
   "tg_tps": 64.64,
   "tg_test": "tg128",
   "commit": "b7552cf",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1637c75133fe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3070 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1689.4,
   "pp_test": "pp512",
   "tg_tps": 63.64,
   "tg_test": "tg128",
   "commit": "ceff6bb",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3070-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ebceec8abfdc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX Vega 64",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 575.06,
   "pp_test": "pp512",
   "tg_tps": 63.43,
   "tg_test": "tg128",
   "commit": "5dcb711",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-vega-64",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "80fdca7ab0c1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 678.14,
   "pp_test": "pp512",
   "tg_tps": 63.16,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "59f82b9dac0b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P102-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 510.62,
   "pp_test": "pp512",
   "tg_tps": 62.83,
   "tg_test": "tg128",
   "commit": "ac2557c",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p102-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c39d2a2db479",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD BC-250",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 370.66,
   "pp_test": "pp512",
   "tg_tps": 62.32,
   "tg_test": "tg128",
   "commit": "5886f4f",
   "contributor": null,
   "notes": null,
   "hardware": "amd-bc-250",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9d6a9356c05d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6650 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1029.52,
   "pp_test": "pp512",
   "tg_tps": 62.14,
   "tg_test": "tg128",
   "commit": "dbb852b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6650-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "76b65bd252a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4060 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2135.66,
   "pp_test": "pp512",
   "tg_tps": 59.53,
   "tg_test": "tg128",
   "commit": "a5c07dc",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4060-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "31ae3e2787a1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 488.06,
   "pp_test": "pp512",
   "tg_tps": 59.36,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a1aee34537cc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GB10",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2789.51,
   "pp_test": "pp512",
   "tg_tps": 58.51,
   "tg_test": "tg128",
   "commit": "657e011",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ade3c3ab9800",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7600",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1167.2,
   "pp_test": "pp512",
   "tg_tps": 58.03,
   "tg_test": "tg128",
   "commit": "91f8c9c",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7600",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0c461ec35bce",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1660 Ti Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 511.67,
   "pp_test": "pp512",
   "tg_tps": 56.6,
   "tg_test": "tg128",
   "commit": "b43556e",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1660-ti-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5dd573d10fe1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Instinct MI25",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 439.42,
   "pp_test": "pp512",
   "tg_tps": 54.69,
   "tg_test": "tg128",
   "commit": "2739a71",
   "contributor": null,
   "notes": null,
   "hardware": "amd-instinct-mi25",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "166997cff059",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6600 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 574.65,
   "pp_test": "pp512",
   "tg_tps": 53.92,
   "tg_test": "tg128",
   "commit": "091592d",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6600-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "94fca8d7bc2d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7600 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 840.85,
   "pp_test": "pp512",
   "tg_tps": 53.02,
   "tg_test": "tg128",
   "commit": "01d8eaa",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7600-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ac868c6e95bf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A770",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1073.85,
   "pp_test": "pp512",
   "tg_tps": 52.56,
   "tg_test": "tg128",
   "commit": "a69d54f",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a770",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ef1773f695b4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro S9300 x2",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 247.26,
   "pp_test": "pp512",
   "tg_tps": 51.86,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": "Split across two GPUs",
   "hardware": "amd-firepro-s9300-x2",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f67209db0e33",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen Al Max+ 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1487.08,
   "pp_test": "pp512",
   "tg_tps": 51.81,
   "tg_test": "tg128",
   "commit": "fb4b273",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b36e8fb2381c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6600",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 758.5,
   "pp_test": "pp512",
   "tg_tps": 50.57,
   "tg_test": "tg128",
   "commit": "86a9c79",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6600",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4abc8f11df94",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX Vega 56",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 439.87,
   "pp_test": "pp512",
   "tg_tps": 50.23,
   "tg_test": "tg128",
   "commit": "92c0b38",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-vega-56",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "167b485dd92e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc B570",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 913.95,
   "pp_test": "pp512",
   "tg_tps": 49.64,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-b570",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9d5cc14c0e0e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3060 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1059.76,
   "pp_test": "pp512",
   "tg_tps": 49.03,
   "tg_test": "tg128",
   "commit": "dbb3a47",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-3060-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "82b2cc4c3f02",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800M",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 861.99,
   "pp_test": "pp512",
   "tg_tps": 48.71,
   "tg_test": "tg128",
   "commit": "8e6f8bc",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bd6f33faf1b8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6600M",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 605.59,
   "pp_test": "pp512",
   "tg_tps": 48.21,
   "tg_test": "tg128",
   "commit": "fe5b78c",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6600m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "01a15335f632",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A770M",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 875.92,
   "pp_test": "pp512",
   "tg_tps": 47.69,
   "tg_test": "tg128",
   "commit": "eeee367",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a770m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dad616682582",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B50",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1186.45,
   "pp_test": "pp512",
   "tg_tps": 46.3,
   "tg_test": "tg128",
   "commit": "aa46bda",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "81ab51dd2e1b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P104-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 311.9,
   "pp_test": "pp512",
   "tg_tps": 46.18,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p104-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "57f7701693cd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A2000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1245.19,
   "pp_test": "pp512",
   "tg_tps": 45.52,
   "tg_test": "tg128",
   "commit": "b1afcab",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ca16977086a0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7600M XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 459.39,
   "pp_test": "pp512",
   "tg_tps": 45.28,
   "tg_test": "tg128",
   "commit": "b9ab0a4",
   "contributor": null,
   "notes": "eGPU",
   "hardware": "amd-radeon-rx-7600m-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9b9c14494968",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro V340",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 375.41,
   "pp_test": "pp512",
   "tg_tps": 45.16,
   "tg_test": "tg128",
   "commit": "9da3dcd",
   "contributor": null,
   "notes": "Split across two GPUs",
   "hardware": "amd-radeon-pro-v340",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c3424dad10ab",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 297.5,
   "pp_test": "pp512",
   "tg_tps": 42.86,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": "eGPU",
   "hardware": "nvidia-gtx-1070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8e940d2c5f3a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A750",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1075.94,
   "pp_test": "pp512",
   "tg_tps": 42.66,
   "tg_test": "tg128",
   "commit": "c1b1876",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a750",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1598a6c7f9ba",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4050 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1154.28,
   "pp_test": "pp512",
   "tg_tps": 41.89,
   "tg_test": "tg128",
   "commit": "d79d8f3",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-4050-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d2a286224d66",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 321.57,
   "pp_test": "pp512",
   "tg_tps": 41.48,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b3f1125be2d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 470",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 207.16,
   "pp_test": "pp512",
   "tg_tps": 40.11,
   "tg_test": "tg128",
   "commit": "32e41fa",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-470",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9f8c3aabc351",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla M40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 92.48,
   "pp_test": "pp512",
   "tg_tps": 39.35,
   "tg_test": "tg128",
   "commit": "b8372ee",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-m40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1a0b3430ade5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 580",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 258.03,
   "pp_test": "pp512",
   "tg_tps": 39.32,
   "tg_test": "tg128",
   "commit": "de4c07f",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-580",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "433207ee1524",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 570",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 226.11,
   "pp_test": "pp512",
   "tg_tps": 37.44,
   "tg_test": "tg128",
   "commit": "e583f3b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-570",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "beaace7642e6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W5500",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 315.39,
   "pp_test": "pp512",
   "tg_tps": 36.82,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-w5500",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "21c920b98d66",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 480",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 248.66,
   "pp_test": "pp512",
   "tg_tps": 34.71,
   "tg_test": "tg128",
   "commit": "3b15924",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-480",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ef04be2145be",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 205.98,
   "pp_test": "pp512",
   "tg_tps": 34.34,
   "tg_test": "tg128",
   "commit": "dbb852b",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2-ultra",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b5676ca4025c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 980",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 186.24,
   "pp_test": "pp512",
   "tg_tps": 33.9,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-980",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4805c6d210e7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P106-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 183.78,
   "pp_test": "pp512",
   "tg_tps": 29.77,
   "tg_test": "tg128",
   "commit": "23bc779",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p106-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "73474c2f104e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro W8100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 155.22,
   "pp_test": "pp512",
   "tg_tps": 29.52,
   "tg_test": "tg128",
   "commit": "4536363",
   "contributor": null,
   "notes": null,
   "hardware": "amd-firepro-w8100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "39aa3030aa4b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P4",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 265.54,
   "pp_test": "pp512",
   "tg_tps": 28.03,
   "tg_test": "tg128",
   "commit": "24d2ee0",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d49c39176f22",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6500 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 255.25,
   "pp_test": "pp512",
   "tg_tps": 27.81,
   "tg_test": "tg128",
   "commit": "g9fdfcd",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6500-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6009b940446f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M3",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 263.7,
   "pp_test": "pp512",
   "tg_tps": 26.39,
   "tg_test": "tg128",
   "commit": "b9ab0a4",
   "contributor": null,
   "notes": "MoltenVK",
   "hardware": "apple-m3",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "760263ceca03",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro S10000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 94.78,
   "pp_test": "pp512",
   "tg_tps": 25.32,
   "tg_test": "tg128",
   "commit": "914a82d",
   "contributor": null,
   "notes": "Split across two GPUs",
   "hardware": "amd-firepro-s10000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7fa0bbacf70f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 1380.07,
   "pp_test": "pp512",
   "tg_tps": 24.45,
   "tg_test": "tg128",
   "commit": "481c65f",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-300-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a16cc7f1ea55",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 200 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 864.99,
   "pp_test": "pp512",
   "tg_tps": 24.37,
   "tg_test": "tg128",
   "commit": "4a89937",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-200-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "16232bef9af5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Quadro P2000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 169.55,
   "pp_test": "pp512",
   "tg_tps": 23.05,
   "tg_test": "tg128",
   "commit": "63f8fe0",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-quadro-p2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cccf817ca753",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen AI 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 479.07,
   "pp_test": "pp512",
   "tg_tps": 22.41,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-300-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "882dc5a85933",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 6000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 240.89,
   "pp_test": "pp512",
   "tg_tps": 21.26,
   "tg_test": "tg128",
   "commit": "ee09828",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-6000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "526dc7b3e4e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 62.7,
   "pp_test": "pp512",
   "tg_tps": 20.95,
   "tg_test": "tg128",
   "commit": "1fe0029",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2-pro",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "406dc3e2dd8e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1050 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 136.42,
   "pp_test": "pp512",
   "tg_tps": 20.96,
   "tg_test": "tg128",
   "commit": "2f0c2db",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1050-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a4f8ff4a281b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 7000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 281.62,
   "pp_test": "pp512",
   "tg_tps": 19.91,
   "tg_test": "tg128",
   "commit": "ebce03e",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-7000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "88ea7e4c2e83",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 8000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 267.27,
   "pp_test": "pp512",
   "tg_tps": 19.81,
   "tg_test": "tg128",
   "commit": "1e9d771",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-8000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7a4f8062b25f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen Z1 Extreme",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 199.36,
   "pp_test": "pp512",
   "tg_tps": 18.77,
   "tg_test": "tg128",
   "commit": "53ff6b9",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-z1-extreme",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "664556740979",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen AI 400 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 131.33,
   "pp_test": "pp512",
   "tg_tps": 17.66,
   "tg_test": "tg128",
   "commit": "cb26014",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-400-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "432d64384c90",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Steam Deck",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 144.31,
   "pp_test": "pp512",
   "tg_tps": 17.52,
   "tg_test": "tg128",
   "commit": "48d22e2",
   "contributor": null,
   "notes": null,
   "hardware": "amd-steam-deck",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f3f377069337",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro D700",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 69.95,
   "pp_test": "pp512",
   "tg_tps": 16.62,
   "tg_test": "tg128",
   "commit": "d3bd719",
   "contributor": null,
   "notes": "MoltenVK, running in FP16 mode on FP32 only chip",
   "hardware": "amd-firepro-d700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d815bf687ba1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro WX 4100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 78.79,
   "pp_test": "pp512",
   "tg_tps": 16.05,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-wx-4100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ebe25e5cae49",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 50.79,
   "pp_test": "pp512",
   "tg_tps": 13.5,
   "tg_test": "tg128",
   "commit": "8c0d6bb",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d97bdd3fa5cf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 38.29,
   "pp_test": "pp512",
   "tg_tps": 12.47,
   "tg_test": "tg128",
   "commit": "2370665",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m1",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b5abbe52779e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 5000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 90.55,
   "pp_test": "pp512",
   "tg_tps": 10.98,
   "tg_test": "tg128",
   "commit": "d84635b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-5000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "26aabcfe2004",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core 1100 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 187.2,
   "pp_test": "pp512",
   "tg_tps": 10.39,
   "tg_test": "tg128",
   "commit": "abb9f3c",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-1100-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "85441d149d27",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 550",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 52.66,
   "pp_test": "pp512",
   "tg_tps": 10.2,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-550",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8b7c863851f0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 4000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 103.87,
   "pp_test": "pp512",
   "tg_tps": 9.63,
   "tg_test": "tg128",
   "commit": "4b385bf",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-4000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7783747178cd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla K80",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 89.46,
   "pp_test": "pp512",
   "tg_tps": 9.39,
   "tg_test": "tg128",
   "commit": "5d46bab",
   "contributor": null,
   "notes": "Running on single GPU",
   "hardware": "nvidia-tesla-k80",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "675e03ac9477",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla K40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 64.37,
   "pp_test": "pp512",
   "tg_tps": 9.3,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-k40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3e10a3e508ac",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "MediaTek Dimensity 9400",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 38.36,
   "pp_test": "pp512",
   "tg_tps": 8.92,
   "tg_test": "tg128",
   "commit": "b9ab0a4",
   "contributor": null,
   "notes": "GPU supports coopmat but pp512 is faster with it turned off",
   "hardware": "mediatek-dimensity-9400",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "494844025e0b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 100 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 185.51,
   "pp_test": "pp512",
   "tg_tps": 8.21,
   "tg_test": "tg128",
   "commit": "1d72c84",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-100-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4ccc0c05a797",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 3000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 48.63,
   "pp_test": "pp512",
   "tg_tps": 8.49,
   "tg_test": "tg128",
   "commit": "1fe0029",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-3000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0f683028b778",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core 1000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 25.58,
   "pp_test": "pp512",
   "tg_tps": 4.25,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-1000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b3bd6ced650e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core 8000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 25.43,
   "pp_test": "pp512",
   "tg_tps": 3.35,
   "tg_test": "tg128",
   "commit": "c4df49a",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-8000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1a4631fb8974",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel N150",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 28.84,
   "pp_test": "pp512",
   "tg_tps": 2.93,
   "tg_test": "tg128",
   "commit": "4f63cd7",
   "contributor": null,
   "notes": null,
   "hardware": "intel-n150",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "345bd3998226",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "CIX CD8180",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": false,
   "pp_tps": 2.8,
   "pp_test": "pp512",
   "tg_tps": 5.51,
   "tg_test": "tg128",
   "commit": "4dca015",
   "contributor": null,
   "notes": null,
   "hardware": "cix-cd8180",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a8217e0371d7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 11796.38,
   "pp_test": "pp512",
   "tg_tps": 273.68,
   "tg_test": "tg128",
   "commit": "ca71fb9",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "429c3e07e27b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5080",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 8054.59,
   "pp_test": "pp512",
   "tg_tps": 192.17,
   "tg_test": "tg128",
   "commit": "f6b533d",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3ba7fae0eee0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 XTX",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3889.54,
   "pp_test": "pp512",
   "tg_tps": 190.92,
   "tg_test": "tg128",
   "commit": "304665f",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b5db66b21610",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 10830.41,
   "pp_test": "pp512",
   "tg_tps": 190.1,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fe951fe7454e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 7567.94,
   "pp_test": "pp512",
   "tg_tps": 176.18,
   "tg_test": "tg128",
   "commit": "0d0764d",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "98cb17c24610",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3090",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 5200.68,
   "pp_test": "pp512",
   "tg_tps": 171.61,
   "tg_test": "tg128",
   "commit": "d05fe1d",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3090",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ee41fb8fc784",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia A100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 7064.4,
   "pp_test": "pp512",
   "tg_tps": 170.56,
   "tg_test": "tg128",
   "commit": "2257758",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-a100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "545549c25714",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5090 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 6011.18,
   "pp_test": "pp512",
   "tg_tps": 154.31,
   "tg_test": "tg128",
   "commit": "4bc272f",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b181744e5987",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon AI Pro R9700",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 5907.66,
   "pp_test": "pp512",
   "tg_tps": 152.7,
   "tg_test": "tg128",
   "commit": "dd1ea52",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-ai-pro-r9700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4ead09688a96",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4080 Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 8007.37,
   "pp_test": "pp512",
   "tg_tps": 150.2,
   "tg_test": "tg128",
   "commit": "81086cd",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6bee23f55d15",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3080",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 4913.83,
   "pp_test": "pp512",
   "tg_tps": 145.74,
   "tg_test": "tg128",
   "commit": "7c7d6ce",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1c348bcdc688",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla V100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1411.25,
   "pp_test": "pp512",
   "tg_tps": 142.13,
   "tg_test": "tg128",
   "commit": "7d77f07",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5b13276381ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A5000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 4071.22,
   "pp_test": "pp512",
   "tg_tps": 140.43,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4bd73d00255e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Ti Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 6801.18,
   "pp_test": "pp512",
   "tg_tps": 135.81,
   "tg_test": "tg128",
   "commit": "4ae88d0",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dff427f2918b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9070 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 5048.07,
   "pp_test": "pp512",
   "tg_tps": 131.54,
   "tg_test": "tg128",
   "commit": "e9fd8dc",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9070-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6c90c959fc42",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7800 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2197.05,
   "pp_test": "pp512",
   "tg_tps": 124.86,
   "tg_test": "tg128",
   "commit": "4fdbc1e",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "42a5cbe8f949",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2701.13,
   "pp_test": "pp512",
   "tg_tps": 120.62,
   "tg_test": "tg128",
   "commit": "71e74a3",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "75a4a8e709fe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2859.98,
   "pp_test": "pp512",
   "tg_tps": 119.51,
   "tg_test": "tg128",
   "commit": "21c17b5",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "50f31161e027",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Instinct MI50",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1127.37,
   "pp_test": "pp512",
   "tg_tps": 117.94,
   "tg_test": "tg128",
   "commit": "3af34b9",
   "contributor": null,
   "notes": null,
   "hardware": "amd-instinct-mi50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "296ec912a6e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia A30",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3158.25,
   "pp_test": "pp512",
   "tg_tps": 117.43,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-a30",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "18adf246cedc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B70",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3150.55,
   "pp_test": "pp512",
   "tg_tps": 114.19,
   "tg_test": "tg128",
   "commit": "b863507",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b70",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "64a89d946d39",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 5515.26,
   "pp_test": "pp512",
   "tg_tps": 113.52,
   "tg_test": "tg128",
   "commit": "516a4ca",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "64fc5c49debc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070 Super",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 5198.91,
   "pp_test": "pp512",
   "tg_tps": 112.0,
   "tg_test": "tg128",
   "commit": "c945aaa",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070-super",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6e6b74be7b66",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon VII",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1007.19,
   "pp_test": "pp512",
   "tg_tps": 110.86,
   "tg_test": "tg128",
   "commit": "77d6ae4",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-vii",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ba679be06da7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7900 GRE",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2251.37,
   "pp_test": "pp512",
   "tg_tps": 110.5,
   "tg_test": "tg128",
   "commit": "4b2a477",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7900-gre",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3f2ec262a633",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Titan V",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 792.74,
   "pp_test": "pp512",
   "tg_tps": 109.21,
   "tg_test": "tg128",
   "commit": "e56abd2",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-titan-v",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8015adce38b7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro VII",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 783.94,
   "pp_test": "pp512",
   "tg_tps": 108.45,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-vii",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f9c7d7293024",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1597.95,
   "pp_test": "pp512",
   "tg_tps": 106.49,
   "tg_test": "tg128",
   "commit": "f1cee99",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1cdf8208db24",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6900 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1761.93,
   "pp_test": "pp512",
   "tg_tps": 106.15,
   "tg_test": "tg128",
   "commit": "a972fae",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3573656b8e7e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 2080 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1936.25,
   "pp_test": "pp512",
   "tg_tps": 100.99,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "80a6899fb6f4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1704.79,
   "pp_test": "pp512",
   "tg_tps": 100.5,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "053aceddd179",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W6800X Duo",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 795.28,
   "pp_test": "pp512",
   "tg_tps": 100.08,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-w6800x-duo",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4bf610203835",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 5060 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3912.65,
   "pp_test": "pp512",
   "tg_tps": 97.01,
   "tg_test": "tg128",
   "commit": "89f10ba",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "99fef9c81848",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 4293.57,
   "pp_test": "pp512",
   "tg_tps": 91.49,
   "tg_test": "tg128",
   "commit": "9a48399",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "770581f4ee31",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3060 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2644.95,
   "pp_test": "pp512",
   "tg_tps": 90.35,
   "tg_test": "tg128",
   "commit": "b42c7fa",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3060-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6a23286dd1ef",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6750 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 997.05,
   "pp_test": "pp512",
   "tg_tps": 82.29,
   "tg_test": "tg128",
   "commit": "228f34c",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6750-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c3dcb00175aa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6700 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1010.9,
   "pp_test": "pp512",
   "tg_tps": 81.86,
   "tg_test": "tg128",
   "commit": "6d75883",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2723792f738e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3060",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2012.88,
   "pp_test": "pp512",
   "tg_tps": 80.59,
   "tg_test": "tg128",
   "commit": "92c0b38",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3060",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "55f39bef0e75",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro V620",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1556.31,
   "pp_test": "pp512",
   "tg_tps": 79.24,
   "tg_test": "tg128",
   "commit": "03d4698",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "012abee35e85",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A4000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2482.74,
   "pp_test": "pp512",
   "tg_tps": 76.07,
   "tg_test": "tg128",
   "commit": "f5245b5",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e98a9f52aa9b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla T10",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1840.14,
   "pp_test": "pp512",
   "tg_tps": 76.05,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-tesla-t10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cb83ef3575c0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 5700 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 538.31,
   "pp_test": "pp512",
   "tg_tps": 74.43,
   "tg_test": "tg128",
   "commit": "4fdbc1e",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-5700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2e1c3ca1e06b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 9060 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3228.15,
   "pp_test": "pp512",
   "tg_tps": 74.01,
   "tg_test": "tg128",
   "commit": "f1cee99",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-9060-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3cd4d68f9772",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc B580",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 419.49,
   "pp_test": "pp512",
   "tg_tps": 72.0,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-b580",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e27104405087",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 557.46,
   "pp_test": "pp512",
   "tg_tps": 71.79,
   "tg_test": "tg128",
   "commit": "1ece0cb6",
   "contributor": null,
   "notes": null,
   "hardware": "apple-m4-max",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dcc063f97dc5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1080 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 647.81,
   "pp_test": "pp512",
   "tg_tps": 71.58,
   "tg_test": "tg128",
   "commit": "ac2557c",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1080-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2e167947ba4a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro W5700",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 446.98,
   "pp_test": "pp512",
   "tg_tps": 71.3,
   "tg_test": "tg128",
   "commit": "23bc779",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-w5700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0774fe87a64e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B60",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 274.76,
   "pp_test": "pp512",
   "tg_tps": 70.54,
   "tg_test": "tg128",
   "commit": "516a4ca",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b60",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f4915d8a5e12",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX Vega 64",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 584.98,
   "pp_test": "pp512",
   "tg_tps": 67.7,
   "tg_test": "tg128",
   "commit": "5dcb711",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-vega-64",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "926cace433cb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 685.51,
   "pp_test": "pp512",
   "tg_tps": 66.48,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3b84fa86a4d5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P102-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 574.56,
   "pp_test": "pp512",
   "tg_tps": 66.48,
   "tg_test": "tg128",
   "commit": "ac2557c",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p102-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0b020e437e04",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD BC-250",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 356.87,
   "pp_test": "pp512",
   "tg_tps": 63.14,
   "tg_test": "tg128",
   "commit": "5886f4f",
   "contributor": null,
   "notes": null,
   "hardware": "amd-bc-250",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "50c643762e23",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 3070 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1832.07,
   "pp_test": "pp512",
   "tg_tps": 62.92,
   "tg_test": "tg128",
   "commit": "ceff6bb",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-3070-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "eee658a0ba01",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6650 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1088.9,
   "pp_test": "pp512",
   "tg_tps": 64.53,
   "tg_test": "tg128",
   "commit": "dbb852b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6650-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "256831fc2f27",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7600",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1353.47,
   "pp_test": "pp512",
   "tg_tps": 60.88,
   "tg_test": "tg128",
   "commit": "91f8c9c",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7600",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8efa7b548493",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GB10",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 3321.91,
   "pp_test": "pp512",
   "tg_tps": 60.39,
   "tg_test": "tg128",
   "commit": "657e011",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "19baf937eae0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX 4060 Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 2358.03,
   "pp_test": "pp512",
   "tg_tps": 60.01,
   "tg_test": "tg128",
   "commit": "a5c07dc",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-4060-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8c69a5507117",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 484.37,
   "pp_test": "pp512",
   "tg_tps": 59.22,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c5b62f12cee7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1660 Ti Mobile",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 514.34,
   "pp_test": "pp512",
   "tg_tps": 57.3,
   "tg_test": "tg128",
   "commit": "b43556e",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1660-ti-laptop",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6d9ad7817b82",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 7600 XT",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1024.38,
   "pp_test": "pp512",
   "tg_tps": 56.11,
   "tg_test": "tg128",
   "commit": "01d8eaa",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-7600-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "59ad486c00b9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro S9300 x2",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 243.33,
   "pp_test": "pp512",
   "tg_tps": 55.64,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": "Split across two GPUs",
   "hardware": "amd-firepro-s9300-x2",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9af822d2d17b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen Al Max+ 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1704.3,
   "pp_test": "pp512",
   "tg_tps": 54.38,
   "tg_test": "tg128",
   "commit": "fb4b273",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3b8e22524b52",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6600",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 804.64,
   "pp_test": "pp512",
   "tg_tps": 53.17,
   "tg_test": "tg128",
   "commit": "86a9c79",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6600",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "6553249d23b8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A770",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1119.68,
   "pp_test": "pp512",
   "tg_tps": 53.07,
   "tg_test": "tg128",
   "commit": "a69d54f",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a770",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a519536b7700",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX Vega 56",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 428.54,
   "pp_test": "pp512",
   "tg_tps": 52.66,
   "tg_test": "tg128",
   "commit": "92c0b38",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-vega-56",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "786e665d9365",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc B570",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 288.51,
   "pp_test": "pp512",
   "tg_tps": 50.49,
   "tg_test": "tg128",
   "commit": "7f76692",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-b570",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a2076b1712f7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 6800M",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 784.16,
   "pp_test": "pp512",
   "tg_tps": 49.06,
   "tg_test": "tg128",
   "commit": "8e6f8bc",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-6800m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3e99031e6005",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia RTX A2000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1361.85,
   "pp_test": "pp512",
   "tg_tps": 45.69,
   "tg_test": "tg128",
   "commit": "b1afcab",
   "contributor": null,
   "notes": "coopmat2",
   "hardware": "nvidia-rtx-a2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "90393b0f24bb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A770M",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 384.74,
   "pp_test": "pp512",
   "tg_tps": 45.68,
   "tg_test": "tg128",
   "commit": "eeee367",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a770m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "7d07c664ddd0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P104-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 325.3,
   "pp_test": "pp512",
   "tg_tps": 48.64,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p104-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e66bdb52c339",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc Pro B50",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1122.53,
   "pp_test": "pp512",
   "tg_tps": 47.69,
   "tg_test": "tg128",
   "commit": "aa46bda",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-pro-b50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "10c56805b0af",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro V340",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 360.23,
   "pp_test": "pp512",
   "tg_tps": 47.54,
   "tg_test": "tg128",
   "commit": "9da3dcd",
   "contributor": null,
   "notes": "Split across two GPUs",
   "hardware": "amd-radeon-pro-v340",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cf94cf334520",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Arc A750",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 303.37,
   "pp_test": "pp512",
   "tg_tps": 43.96,
   "tg_test": "tg128",
   "commit": "c1b1876",
   "contributor": null,
   "notes": null,
   "hardware": "intel-arc-a750",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0afb1b2642bf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1070 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 292.85,
   "pp_test": "pp512",
   "tg_tps": 43.42,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": "eGPU",
   "hardware": "nvidia-gtx-1070-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c6514be9fbfe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 470",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 211.25,
   "pp_test": "pp512",
   "tg_tps": 43.35,
   "tg_test": "tg128",
   "commit": "32e41fa",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-470",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "51fbc89e5ec8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1070",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 330.84,
   "pp_test": "pp512",
   "tg_tps": 43.33,
   "tg_test": "tg128",
   "commit": "360d653",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e0117b58ad19",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla M40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 93.35,
   "pp_test": "pp512",
   "tg_tps": 41.68,
   "tg_test": "tg128",
   "commit": "b8372ee",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-m40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "69a8dff91838",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 570",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 229.23,
   "pp_test": "pp512",
   "tg_tps": 40.18,
   "tg_test": "tg128",
   "commit": "e583f3b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-570",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d8d5fac9f813",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon RX 480",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 194.52,
   "pp_test": "pp512",
   "tg_tps": 37.23,
   "tg_test": "tg128",
   "commit": "0bcb40b",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-rx-480",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e14aee8c7715",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 198.83,
   "pp_test": "pp512",
   "tg_tps": 36.07,
   "tg_test": "tg128",
   "commit": "dbb852b",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2-ultra",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "64bbab62b1be",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 980",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 180.97,
   "pp_test": "pp512",
   "tg_tps": 34.16,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-980",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "505fa2ccde79",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia P106-100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 183.4,
   "pp_test": "pp512",
   "tg_tps": 30.79,
   "tg_test": "tg128",
   "commit": "23bc779",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-p106-100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "31b88fbc76eb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD FirePro W8100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 140.52,
   "pp_test": "pp512",
   "tg_tps": 29.28,
   "tg_test": "tg128",
   "commit": "4536363",
   "contributor": null,
   "notes": null,
   "hardware": "amd-firepro-w8100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bb5871431990",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla P4",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 287.14,
   "pp_test": "pp512",
   "tg_tps": 28.37,
   "tg_test": "tg128",
   "commit": "24d2ee0",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-p4",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ce63bfb7e38d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 200 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 867.27,
   "pp_test": "pp512",
   "tg_tps": 25.12,
   "tg_test": "tg128",
   "commit": "4a89937",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-200-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c0d04e41fe8d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 1332.41,
   "pp_test": "pp512",
   "tg_tps": 23.93,
   "tg_test": "tg128",
   "commit": "481c65f",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-300-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "90df0463558f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Quadro P2000",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 181.71,
   "pp_test": "pp512",
   "tg_tps": 23.77,
   "tg_test": "tg128",
   "commit": "63f8fe0",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-quadro-p2000",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bb576eb6c5c3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen AI 300 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 532.59,
   "pp_test": "pp512",
   "tg_tps": 23.93,
   "tg_test": "tg128",
   "commit": "N/A",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-300-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "112c456a5515",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 8000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 311.02,
   "pp_test": "pp512",
   "tg_tps": 21.28,
   "tg_test": "tg128",
   "commit": "1e9d771",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-8000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bfa02f31b36a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 6000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 277.91,
   "pp_test": "pp512",
   "tg_tps": 21.15,
   "tg_test": "tg128",
   "commit": "ee09828",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-6000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "74c59825a153",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2 Pro",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 58.86,
   "pp_test": "pp512",
   "tg_tps": 20.97,
   "tg_test": "tg128",
   "commit": "1fe0029",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2-pro",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "29a98bbb5765",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 7000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 312.85,
   "pp_test": "pp512",
   "tg_tps": 20.09,
   "tg_test": "tg128",
   "commit": "835b2b9",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-7000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "804f0ed36cd6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia GTX 1050 Ti",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 127.54,
   "pp_test": "pp512",
   "tg_tps": 20.08,
   "tg_test": "tg128",
   "commit": "2f0c2db",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-gtx-1050-ti",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "683d493147ea",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen AI 400 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 150.12,
   "pp_test": "pp512",
   "tg_tps": 19.03,
   "tg_test": "tg128",
   "commit": "cb26014",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-ai-400-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "660e9265ef40",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Steam Deck",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 164.34,
   "pp_test": "pp512",
   "tg_tps": 18.4,
   "tg_test": "tg128",
   "commit": "48d22e2",
   "contributor": null,
   "notes": null,
   "hardware": "amd-steam-deck",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "54271dba871f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Radeon Pro WX 4100",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 75.59,
   "pp_test": "pp512",
   "tg_tps": 16.56,
   "tg_test": "tg128",
   "commit": "860a9e4",
   "contributor": null,
   "notes": null,
   "hardware": "amd-radeon-pro-wx-4100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "835240990906",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M1",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 35.93,
   "pp_test": "pp512",
   "tg_tps": 12.85,
   "tg_test": "tg128",
   "commit": "2370665",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m1",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e9db65045cc8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Apple M2",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 46.81,
   "pp_test": "pp512",
   "tg_tps": 12.25,
   "tg_test": "tg128",
   "commit": "8c0d6bb",
   "contributor": null,
   "notes": "Asahi Linux",
   "hardware": "apple-m2",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bc81e95e2891",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 5000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 79.06,
   "pp_test": "pp512",
   "tg_tps": 10.75,
   "tg_test": "tg128",
   "commit": "5d195f1",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-5000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "37807c482749",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core 1100 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 174.77,
   "pp_test": "pp512",
   "tg_tps": 10.58,
   "tg_test": "tg128",
   "commit": "abb9f3c",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-1100-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2220fb5392c5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla K40",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 64.37,
   "pp_test": "pp512",
   "tg_tps": 9.92,
   "tg_test": "tg128",
   "commit": "eec1e33",
   "contributor": null,
   "notes": null,
   "hardware": "nvidia-tesla-k40",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "884ef974a62d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 4000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 113.32,
   "pp_test": "pp512",
   "tg_tps": 9.87,
   "tg_test": "tg128",
   "commit": "4b385bf",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-4000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e6533c75ea2e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Nvidia Tesla K80",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 88.26,
   "pp_test": "pp512",
   "tg_tps": 9.49,
   "tg_test": "tg128",
   "commit": "5d46bab",
   "contributor": null,
   "notes": "Running on single GPU",
   "hardware": "nvidia-tesla-k80",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e38baca7c22f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AMD Ryzen 5 3000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 47.41,
   "pp_test": "pp512",
   "tg_tps": 8.47,
   "tg_test": "tg128",
   "commit": "1fe0029",
   "contributor": null,
   "notes": null,
   "hardware": "amd-ryzen-5-3000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0a315ba771a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core Ultra 100 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 77.66,
   "pp_test": "pp512",
   "tg_tps": 7.75,
   "tg_test": "tg128",
   "commit": "2e89f76",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-ultra-100-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "69c9453fae01",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel Core 8000 Series",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 25.55,
   "pp_test": "pp512",
   "tg_tps": 3.35,
   "tg_test": "tg128",
   "commit": "c4df49a",
   "contributor": null,
   "notes": null,
   "hardware": "intel-core-8000-series",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ab8a4c382041",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-10879",
   "source_name": "llama.cpp Vulkan scoreboard (#10879)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/10879",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Intel N150",
   "quant": "Q4_0",
   "backend": "Vulkan",
   "flash_attn": true,
   "pp_tps": 25.59,
   "pp_test": "pp512",
   "tg_tps": 2.91,
   "tg_test": "tg128",
   "commit": "4f63cd7",
   "contributor": null,
   "notes": null,
   "hardware": "intel-n150",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dc0373ff85cd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI300X",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 11476.4,
   "pp_test": "pp512",
   "tg_tps": 232.92,
   "tg_test": "tg128",
   "commit": "ee3a9fc",
   "contributor": "@yeahdongcn",
   "notes": null,
   "hardware": "amd-instinct-mi300x",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e641a008b003",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 XTX",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 3552.27,
   "pp_test": "pp512",
   "tg_tps": 167.11,
   "tg_test": "tg128",
   "commit": "2f0c2db",
   "contributor": "@Diablo-D3",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "27df35b1d714",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI210",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 2486.22,
   "pp_test": "pp512",
   "tg_tps": 124.51,
   "tg_test": "tg128",
   "commit": "8160b38",
   "contributor": "@65a",
   "notes": null,
   "hardware": "amd-instinct-mi210",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "85796f051f02",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Pro W7900",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 3213.17,
   "pp_test": "pp512",
   "tg_tps": 121.18,
   "tg_test": "tg128",
   "commit": "8160b38",
   "contributor": "@65a",
   "notes": null,
   "hardware": "amd-radeon-pro-w7900",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d7321a04eb1f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 3098.38,
   "pp_test": "pp512",
   "tg_tps": 116.15,
   "tg_test": "tg128",
   "commit": "1e15bfd",
   "contributor": "@AdamNiederer",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "375fb34e3c04",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9070",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 2381.77,
   "pp_test": "pp512",
   "tg_tps": 114.48,
   "tg_test": "tg128",
   "commit": "d0660f2",
   "contributor": "@andj1210",
   "notes": null,
   "hardware": "amd-radeon-rx-9070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bcd4d0f7a001",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI100",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 2732.83,
   "pp_test": "pp512",
   "tg_tps": 110.48,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@firefox42",
   "notes": null,
   "hardware": "amd-instinct-mi100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9640a075add3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9070 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 5055.19,
   "pp_test": "pp512",
   "tg_tps": 101.27,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": "@Hadrianneue",
   "notes": null,
   "hardware": "amd-radeon-rx-9070-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2a65612288c6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7800 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 2151.81,
   "pp_test": "pp512",
   "tg_tps": 100.94,
   "tg_test": "tg128",
   "commit": "00131d6",
   "contributor": "@olegshulyakov",
   "notes": null,
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9f5b63d0261b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI50",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1057.24,
   "pp_test": "pp512",
   "tg_tps": 98.95,
   "tg_test": "tg128",
   "commit": "97d5117",
   "contributor": "@wtarreau",
   "notes": null,
   "hardware": "amd-instinct-mi50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c10fd515775f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 GRE",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1456.98,
   "pp_test": "pp512",
   "tg_tps": 96.07,
   "tg_test": "tg128",
   "commit": "6fa3b55",
   "contributor": "@MihaiBojescu",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-gre",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c6ad30b9485b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AI PRO R9700",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 4443.54,
   "pp_test": "pp512",
   "tg_tps": 93.84,
   "tg_test": "tg128",
   "commit": "bd4ef13",
   "contributor": "@gogich77",
   "notes": null,
   "hardware": "amd-radeon-ai-pro-r9700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "27f8124f1094",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI60",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1289.11,
   "pp_test": "pp512",
   "tg_tps": 91.46,
   "tg_test": "tg128",
   "commit": "504af20",
   "contributor": "@Said-Akbar",
   "notes": null,
   "hardware": "amd-instinct-mi60",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2f3fc7460e84",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 6900 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1889.84,
   "pp_test": "pp512",
   "tg_tps": 88.49,
   "tg_test": "tg128",
   "commit": "a972fae",
   "contributor": "@notgood",
   "notes": null,
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "44b7db240885",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Pro VII",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1064.99,
   "pp_test": "pp512",
   "tg_tps": 87.45,
   "tg_test": "tg128",
   "commit": "2739a71",
   "contributor": "@8XXD8",
   "notes": null,
   "hardware": "amd-radeon-pro-vii",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5edbd5c4c038",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 6800 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1447.07,
   "pp_test": "pp512",
   "tg_tps": 83.92,
   "tg_test": "tg128",
   "commit": "79c1160",
   "contributor": "@MrLavender",
   "notes": null,
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "557df0b9bd2f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Pro V620",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1803.65,
   "pp_test": "pp512",
   "tg_tps": 74.66,
   "tg_test": "tg128",
   "commit": "5c0eb5e",
   "contributor": "@samteezy",
   "notes": null,
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4c4df78462dc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9060 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1419.67,
   "pp_test": "pp512",
   "tg_tps": 67.58,
   "tg_test": "tg128",
   "commit": "a0e13dc",
   "contributor": "@lcy0321",
   "notes": null,
   "hardware": "amd-radeon-rx-9060-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bd4ce7b161dd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 5700 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 354.17,
   "pp_test": "pp512",
   "tg_tps": 67.55,
   "tg_test": "tg128",
   "commit": "c05e8c9",
   "contributor": "@daniandtheweb",
   "notes": null,
   "hardware": "amd-radeon-rx-5700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "cd761c11b36e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI25",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 409.83,
   "pp_test": "pp512",
   "tg_tps": 63.94,
   "tg_test": "tg128",
   "commit": "2739a71",
   "contributor": "@8XXD8",
   "notes": null,
   "hardware": "amd-instinct-mi25",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "53aa9cca9da2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AI Max+ 395",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 911.36,
   "pp_test": "pp512",
   "tg_tps": 50.01,
   "tg_test": "tg128",
   "commit": "e60f241",
   "contributor": "@firefox42",
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b0a9ebf2b802",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7600 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 1099.64,
   "pp_test": "pp512",
   "tg_tps": 48.58,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@wbruna",
   "notes": null,
   "hardware": "amd-radeon-rx-7600-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1d2f224c8967",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX Vega 64",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 240.68,
   "pp_test": "pp512",
   "tg_tps": 48.46,
   "tg_test": "tg128",
   "commit": "ec428b0",
   "contributor": "@davispuh",
   "notes": null,
   "hardware": "amd-radeon-rx-vega-64",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "57236ad18364",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Radeon 8060S",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 351.36,
   "pp_test": "pp512",
   "tg_tps": 47.97,
   "tg_test": "tg128",
   "commit": "1d0125b",
   "contributor": "@hspak",
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e962ae1ed577",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Radeon 880M",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": false,
   "pp_tps": 163.25,
   "pp_test": "pp512",
   "tg_tps": 12.97,
   "tg_test": "tg128",
   "commit": "c55d53a",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "amd-radeon-880m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "00bdb7bd2337",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI300X",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 11945.97,
   "pp_test": "pp512",
   "tg_tps": 218.53,
   "tg_test": "tg128",
   "commit": "ee3a9fc",
   "contributor": "@yeahdongcn",
   "notes": null,
   "hardware": "amd-instinct-mi300x",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3a63deab0726",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 XTX",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 3874.25,
   "pp_test": "pp512",
   "tg_tps": 170.12,
   "tg_test": "tg128",
   "commit": "2f0c2db",
   "contributor": "@Diablo-D3",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "767ace8a5f50",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI210",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 2571.82,
   "pp_test": "pp512",
   "tg_tps": 130.18,
   "tg_test": "tg128",
   "commit": "8160b38",
   "contributor": "@65a",
   "notes": null,
   "hardware": "amd-instinct-mi210",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5567573c4d70",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Pro W7900",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 3472.86,
   "pp_test": "pp512",
   "tg_tps": 127.43,
   "tg_test": "tg128",
   "commit": "8160b38",
   "contributor": "@65a",
   "notes": null,
   "hardware": "amd-radeon-pro-w7900",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a6a632067bb2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9070",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 2452.68,
   "pp_test": "pp512",
   "tg_tps": 115.32,
   "tg_test": "tg128",
   "commit": "d0660f2",
   "contributor": "@andj1210",
   "notes": null,
   "hardware": "amd-radeon-rx-9070",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "b43aa8c91e77",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 3261.75,
   "pp_test": "pp512",
   "tg_tps": 112.3,
   "tg_test": "tg128",
   "commit": "1e15bfd",
   "contributor": "@AdamNiederer",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2a60a3c38ba7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI100",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 2755.0,
   "pp_test": "pp512",
   "tg_tps": 104.71,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@firefox42",
   "notes": null,
   "hardware": "amd-instinct-mi100",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "a40d59aa7a1c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Instinct MI50",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1129.43,
   "pp_test": "pp512",
   "tg_tps": 105.82,
   "tg_test": "tg128",
   "commit": "97d5117",
   "contributor": "@wtarreau",
   "notes": null,
   "hardware": "amd-instinct-mi50",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fda94f7bfc3b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AI PRO R9700",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 4773.07,
   "pp_test": "pp512",
   "tg_tps": 97.98,
   "tg_test": "tg128",
   "commit": "bd4ef13",
   "contributor": "@gogich77",
   "notes": null,
   "hardware": "amd-radeon-ai-pro-r9700",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "df53c6c909a0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7900 GRE",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1598.79,
   "pp_test": "pp512",
   "tg_tps": 97.53,
   "tg_test": "tg128",
   "commit": "6fa3b55",
   "contributor": "@MihaiBojescu",
   "notes": null,
   "hardware": "amd-radeon-rx-7900-gre",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "dd5cb7253dc0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9070 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 4903.51,
   "pp_test": "pp512",
   "tg_tps": 97.28,
   "tg_test": "tg128",
   "commit": "583cb83",
   "contributor": "@Hadrianneue",
   "notes": null,
   "hardware": "amd-radeon-rx-9070-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d3f4b7f5b5fe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7800 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 2304.63,
   "pp_test": "pp512",
   "tg_tps": 95.99,
   "tg_test": "tg128",
   "commit": "00131d6",
   "contributor": "@olegshulyakov",
   "notes": null,
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "992c3cf5e1da",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 6900 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1948.31,
   "pp_test": "pp512",
   "tg_tps": 85.04,
   "tg_test": "tg128",
   "commit": "a972fae",
   "contributor": "@notgood",
   "notes": null,
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1bd204d05f3c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Pro V620",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1256.86,
   "pp_test": "pp512",
   "tg_tps": 70.83,
   "tg_test": "tg128",
   "commit": "5c0eb5e",
   "contributor": "@samteezy",
   "notes": null,
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1b6716f56d2d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 9060 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1479.27,
   "pp_test": "pp512",
   "tg_tps": 65.42,
   "tg_test": "tg128",
   "commit": "a0e13dc",
   "contributor": "@lcy0321",
   "notes": null,
   "hardware": "amd-radeon-rx-9060-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "deb99c8261c1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 5700 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 314.17,
   "pp_test": "pp512",
   "tg_tps": 62.02,
   "tg_test": "tg128",
   "commit": "c05e8c9",
   "contributor": "@daniandtheweb",
   "notes": null,
   "hardware": "amd-radeon-rx-5700-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "78536772035e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "AI Max+ 395",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1003.53,
   "pp_test": "pp512",
   "tg_tps": 49.87,
   "tg_test": "tg128",
   "commit": "e60f241",
   "contributor": "@firefox42",
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "48773f48e250",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Radeon 8060S",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 366.08,
   "pp_test": "pp512",
   "tg_tps": 48.97,
   "tg_test": "tg128",
   "commit": "1d0125b",
   "contributor": "@hspak",
   "notes": null,
   "hardware": "amd-ryzen-ai-maxplus-395-radeon-8060s",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "be3742cf0d39",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX 7600 XT",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 1199.16,
   "pp_test": "pp512",
   "tg_tps": 47.65,
   "tg_test": "tg128",
   "commit": "9c35706",
   "contributor": "@wbruna",
   "notes": null,
   "hardware": "amd-radeon-rx-7600-xt",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ccec9bba6de1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "RX Vega 64",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 153.17,
   "pp_test": "pp512",
   "tg_tps": 42.46,
   "tg_test": "tg128",
   "commit": "ec428b0",
   "contributor": "@davispuh",
   "notes": null,
   "hardware": "amd-radeon-rx-vega-64",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "41025843adc7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15021",
   "source_name": "llama.cpp ROCm scoreboard (#15021)",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15021",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:35Z",
   "hardware_raw": "Radeon 880M",
   "quant": "Q4_0",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 213.31,
   "pp_test": "pp512",
   "tg_tps": 16.16,
   "tg_test": "tg128",
   "commit": "c55d53a",
   "contributor": "@Hedede",
   "notes": null,
   "hardware": "amd-radeon-880m",
   "model": "llama-2-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f878b2dd1886",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1704",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7594.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 119.61,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 1/422",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2964c8400a9b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/673",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6807.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 120.98,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 2/422",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b47918363900",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/77",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6696.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 91.03,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 3/422",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5c69759b4165",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/937",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6345.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 85.77,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 4/422",
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8a68faa5f74e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/69",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5202.16,
   "pp_test": "LocalScore avg",
   "tg_tps": 119.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 5/422",
   "hardware": "nvidia-h100-pcie",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d3c2c2ec2993",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/155",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6297.37,
   "pp_test": "LocalScore avg",
   "tg_tps": 66.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 6/422",
   "hardware": "nvidia-rtx-5090",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9090b1773f99",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4016",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4462.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 105.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 7/422",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3631d487ac46",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/15",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4791.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 88.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 8/422",
   "hardware": "nvidia-rtx-4080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "684c7d1618a0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2949",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4978.68,
   "pp_test": "LocalScore avg",
   "tg_tps": 78.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 9/422",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bfc4affa7842",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/127",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4023.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 109.78,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 10/422",
   "hardware": "nvidia-rtx-3090-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "669f95ee54bd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/61",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100-SXM4-80GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3570.51,
   "pp_test": "LocalScore avg",
   "tg_tps": 110.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 11/422",
   "hardware": "nvidia-a100-80gb-sxm",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "eea9f45f3b1f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1344",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5543.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 12/422",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b8f462980b0e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/86",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5487.5,
   "pp_test": "LocalScore avg",
   "tg_tps": 51.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 13/422",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7cd24f511407",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3553",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3829.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 90.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 14/422",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e5caca1526b6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/84",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5327.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 15/422",
   "hardware": "nvidia-l40s",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a5505a9866fa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3549.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 96.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 16/422",
   "hardware": "nvidia-rtx-3090",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a2ae03bc75a7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/120",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3514.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 90.49,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 17/422",
   "hardware": "nvidia-rtx-3080-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7974a3e87b7d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/82",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5064.74,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.15,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 18/422",
   "hardware": "nvidia-l40",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9bb9d6743914",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/50",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3447.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 90.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 19/422",
   "hardware": "nvidia-rtx-a6000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "90eb5220d648",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/102",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4461.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 54.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 20/422",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6dc601df2c9b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/87",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 80GB HBM3",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4686.8,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 21/422",
   "hardware": "nvidia-h100-sxm",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e073e509b992",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/489",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4427.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 44.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 22/422",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9aed5eb515a6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/160",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3656.83,
   "pp_test": "LocalScore avg",
   "tg_tps": 65.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 23/422",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8d76dc03e075",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/138",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3768.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 24/422",
   "hardware": "nvidia-rtx-4070-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d72c1878c532",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/49",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3270.17,
   "pp_test": "LocalScore avg",
   "tg_tps": 80.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 25/422",
   "hardware": "nvidia-a40",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9b2d51277132",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/147",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3192.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 76.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 26/422",
   "hardware": "nvidia-rtx-4070",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7f36f71a6ae8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2948",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3030.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 81.65,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 27/422",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "79444b69d720",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/26",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2890.51,
   "pp_test": "LocalScore avg",
   "tg_tps": 89.05,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 28/422",
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cc836cc38378",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/132",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3018.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 80.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 29/422",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "470f463775ec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/78",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 NVL",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4298.75,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 30/422",
   "hardware": "nvidia-h100-nvl",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b52f82cdcf5c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/136",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3561.02,
   "pp_test": "LocalScore avg",
   "tg_tps": 53.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 31/422",
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9fe7a45d0764",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/315",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3335.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 53.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 32/422",
   "hardware": "nvidia-rtx-4070-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9c6a8a25be27",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3548",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2874.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 65.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 33/422",
   "hardware": "nvidia-rtx-5070-ti-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c8bc27633b79",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/122",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2509.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 83.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 34/422",
   "hardware": "nvidia-rtx-3070-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3ea83e5f6877",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/775",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2256.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 90.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 35/422",
   "hardware": "nvidia-tesla-v100-16gb",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a9b2e26915c7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/51",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4500",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2498.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 77.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 36/422",
   "hardware": "nvidia-rtx-a4500",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0c2e0d9ca962",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/168",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2927.06,
   "pp_test": "LocalScore avg",
   "tg_tps": 55.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 37/422",
   "hardware": "nvidia-rtx-5070",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9de30fb0218f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/920",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-32GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2176.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 88.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 38/422",
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9a81670818e9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3817",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN RTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2226.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 71.22,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 39/422",
   "hardware": "nvidia-titan-rtx",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6aa6c34eddea",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1119",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3662.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 26.11,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 40/422",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4855a603510b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/76",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2608.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 53.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 41/422",
   "hardware": "nvidia-rtx-4000-ada",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ef29704d0884",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/112",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2026.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 72.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 42/422",
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "105dd2dce141",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/860",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2364.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 59.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 43/422",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7326ca685ec6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/107",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN V",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1822.65,
   "pp_test": "LocalScore avg",
   "tg_tps": 73.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 44/422",
   "hardware": "nvidia-titan-v",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ad4f1c9c7a5d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4230",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2030.2,
   "pp_test": "LocalScore avg",
   "tg_tps": 57.68,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 45/422",
   "hardware": "nvidia-rtx-5070-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8eb84d14f8d9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3523",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1891.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.86,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 46/422",
   "hardware": "nvidia-quadro-rtx-6000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5c63dac1e01d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/64",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100 80GB PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2600.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 47/422",
   "hardware": "nvidia-a100-80gb-pcie",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0e4bbfc5f0de",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/632",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 3500 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2444.97,
   "pp_test": "LocalScore avg",
   "tg_tps": 41.79,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 48/422",
   "hardware": "nvidia-rtx-3500-ada-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3f1ce7b6b28b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1847",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2574.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 49/422",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fce769258134",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/126",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2005.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 59.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 50/422",
   "hardware": "nvidia-rtx-3070",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "85670d9dd34f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/7",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2233.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 48.32,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 51/422",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f2939980afe7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/741",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1909.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 62.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 52/422",
   "hardware": "nvidia-rtx-3070-ti-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b9feb8a1b50f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2029",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2052.3,
   "pp_test": "LocalScore avg",
   "tg_tps": 53.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 53/422",
   "hardware": "nvidia-rtx-3080-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8d528f15cfff",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/44",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1930.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 56.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 54/422",
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bd71fa96ab58",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/121",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1839.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 60.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 55/422",
   "hardware": "nvidia-rtx-3060-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "545a82b77bbc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2298",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2313.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 51.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 56/422",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1c4b280997f9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/894",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1884.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 52.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 57/422",
   "hardware": "nvidia-rtx-3080-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "aec8b31aedef",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1537",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2130.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 58/422",
   "hardware": "nvidia-rtx-4070-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "849bef1e0cac",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2011",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1693.29,
   "pp_test": "LocalScore avg",
   "tg_tps": 59.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 59/422",
   "hardware": "nvidia-rtx-2080-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a31cac6b67dd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/63",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L4",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2056.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 60/422",
   "hardware": "nvidia-l4",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0c52b50bfe0e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/157",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 SFF Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2003.96,
   "pp_test": "LocalScore avg",
   "tg_tps": 44.37,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 61/422",
   "hardware": "nvidia-rtx-4000-sff-ada",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2b6a4e3b7424",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/328",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2086.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 38.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 62/422",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6a0496de865a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/363",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2085.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 63/422",
   "hardware": "nvidia-a10",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d5abcd5eb0b1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/569",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1677.26,
   "pp_test": "LocalScore avg",
   "tg_tps": 74.02,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 64/422",
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "53af5f6ef9ba",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1476",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1545.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 58.43,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 65/422",
   "hardware": "nvidia-rtx-2080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4b94771eec48",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2944",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2009.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 66/422",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f2ae384b45bf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1771",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1986.31,
   "pp_test": "LocalScore avg",
   "tg_tps": 33.63,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 67/422",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2e8db0f0d186",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/101",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2070 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1443.97,
   "pp_test": "LocalScore avg",
   "tg_tps": 57.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 68/422",
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fc1c8feef207",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/43",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1482.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 51.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 69/422",
   "hardware": "nvidia-rtx-3060",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ffa493fb05e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1024",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1497.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 70/422",
   "hardware": "nvidia-rtx-a5000-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e6aad2f5715a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/137",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1544.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 38.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 71/422",
   "hardware": "nvidia-rtx-4060",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "749471ffe392",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/404",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1109.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 62.68,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 72/422",
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a1d1f8a50ba3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1727",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1251.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 52.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 73/422",
   "hardware": "nvidia-rtx-2070",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a1a104038f63",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1487",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 2000 Blackwell Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1453.06,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 74/422",
   "hardware": "nvidia-rtx-pro-2000-blackwell-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0c34f8ac7666",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1359",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1062.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 63.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 75/422",
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d1d2316724bd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/111",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2060 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1205.22,
   "pp_test": "LocalScore avg",
   "tg_tps": 49.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 76/422",
   "hardware": "nvidia-rtx-2060-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8f2fadcadb07",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/62",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 2000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1356.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.55,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 77/422",
   "hardware": "nvidia-rtx-2000-ada",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e3d2ed99e74a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4053",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1339.87,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 78/422",
   "hardware": "nvidia-rtx-3060",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "807509dedcc2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3572",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1175.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 41.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 79/422",
   "hardware": "nvidia-rtx-3070-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "146d68dd2e17",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/767",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1252.83,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 80/422",
   "hardware": "nvidia-rtx-3060-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b825c2cd58fd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/529",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1365.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 81/422",
   "hardware": "nvidia-rtx-4060-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bf42000935a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/319",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6900 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1082.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.42,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 82/422",
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3152e9b82cbb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1463",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 4000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 992.39,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.0,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 83/422",
   "hardware": "nvidia-quadro-rtx-4000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9a2f1ea6fe75",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/598",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4050 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1200.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 31.36,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 84/422",
   "hardware": "nvidia-rtx-4050-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5dfe7f4c42bd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4247",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 718.42,
   "pp_test": "LocalScore avg",
   "tg_tps": 71.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 85/422",
   "hardware": "apple-m5-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "866fa68bcf2e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/880",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 885.1,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 86/422",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "31464198c926",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/665",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 960.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 38.08,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 87/422",
   "hardware": "nvidia-rtx-2060",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bee91077ba7f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3256",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 2000 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1150.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 29.51,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 88/422",
   "hardware": "nvidia-rtx-2000-ada-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8e632b5ae7a0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/514",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 12GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 976.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.73,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 89/422",
   "hardware": "nvidia-rtx-a2000-12gb",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cbe72ff8628b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2309",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 705.74,
   "pp_test": "LocalScore avg",
   "tg_tps": 59.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 90/422",
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "93866e167a30",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2294",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 979.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 91/422",
   "hardware": "nvidia-rtx-3050",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d427f3e9fcb8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3339",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 OEM",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 964.79,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.96,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 92/422",
   "hardware": "nvidia-rtx-3050",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e17ebe11d7a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/475",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 950.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 93/422",
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4cfdc276152e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1087",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+64GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 667.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 54.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 94/422",
   "hardware": "apple-m1-ultra-64c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9685bf8d5554",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3105",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 637.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 62.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 95/422",
   "hardware": "apple-m5-max-32c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d39199bbd361",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/491",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 864.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 39.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 96/422",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fc22b4035d21",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1188",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 656.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 54.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 97/422",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6a58545b5836",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/6",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 631.76,
   "pp_test": "LocalScore avg",
   "tg_tps": 52.85,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 98/422",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e777f52cd3b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3883",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 1000 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 954.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.18,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 99/422",
   "hardware": "nvidia-rtx-1000-ada-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "216e2644045f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/776",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 565.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 47.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 100/422",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cd5365b1ba69",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/731",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 8GB Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 790.31,
   "pp_test": "LocalScore avg",
   "tg_tps": 27.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 101/422",
   "hardware": "nvidia-rtx-a2000-8gb-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f42021d14de9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/750",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 582.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 102/422",
   "hardware": "apple-m3-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bc3ed6a80bec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/306",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+48GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 534.35,
   "pp_test": "LocalScore avg",
   "tg_tps": 48.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 103/422",
   "hardware": "apple-m1-ultra-48c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4c806d002863",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/494",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 737.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 30.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 104/422",
   "hardware": "amd-radeon-rx-6800",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1be0b5fdbb05",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1998",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 6GB Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 789.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 105/422",
   "hardware": "nvidia-rtx-3050-6gb-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "04ea892d31d3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2077",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 1000 Blackwell Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 906.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.85,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 106/422",
   "hardware": "nvidia-rtx-pro-1000-blackwell-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4d1b9dda5e7e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/162",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 10P+4E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 540.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 107/422",
   "hardware": "apple-m4-max-32c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "392d2543154d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1425",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 758.3,
   "pp_test": "LocalScore avg",
   "tg_tps": 28.73,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 108/422",
   "hardware": "nvidia-rtx-a2000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a9ffd38dde84",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/977",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 778.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 109/422",
   "hardware": "nvidia-rtx-3050",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c457c97c115c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/515",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 484.06,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 110/422",
   "hardware": "apple-m2-max-38c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4a9f13d65aae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1003",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD  Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 524.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 38.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 111/422",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8e62c3e4c3c8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1291",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 473.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 44.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 112/422",
   "hardware": "apple-m2-max-38c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9769b1affbdd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3949",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 3000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 606.47,
   "pp_test": "LocalScore avg",
   "tg_tps": 29.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 113/422",
   "hardware": "nvidia-quadro-rtx-3000",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "25459bacbdd6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/110",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 652.37,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 114/422",
   "hardware": "nvidia-gtx-1080",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "eeaab674ed1a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/789",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 456.81,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.74,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 115/422",
   "hardware": "apple-m3-max-30c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "23ea259ffe52",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3180",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 411.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 44.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 116/422",
   "hardware": "apple-m5-pro-20c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "123ab7a96cb6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/805",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 443.5,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 117/422",
   "hardware": "apple-m3-max-30c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e5460aa834a7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3226",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 410.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 44.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 118/422",
   "hardware": "apple-m5-pro-20c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4e64bdab6bfd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/624",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6600 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 604.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 119/422",
   "hardware": "amd-radeon-rx-6600-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "21c21b6e9002",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/100",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 427.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 33.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 120/422",
   "hardware": "nvidia-tesla-p100",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "155fd07c7e3f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/12",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6650 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 535.6,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 121/422",
   "hardware": "amd-radeon-rx-6650-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8b7abcd9348e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/364",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 372.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 122/422",
   "hardware": "apple-m1-max-32c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3384047b7370",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1664",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6950 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1023.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 6.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 123/422",
   "hardware": "amd-radeon-rx-6950-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2b7d01059987",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3323",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 5P+10E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 340.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 43.74,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 124/422",
   "hardware": "apple-m5-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9725f9febe0e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/607",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 402.68,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 125/422",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8a746092ab2b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3531",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 5P+10E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 336.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 43.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 126/422",
   "hardware": "apple-m5-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b9aadd899fca",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/597",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 365.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.85,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 127/422",
   "hardware": "apple-m1-max-32c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a06f6c8c7dc0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1079",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 362.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 128/422",
   "hardware": "apple-m2-max-30c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "79e76c14f3ff",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3387",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-PCIE-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 371.51,
   "pp_test": "LocalScore avg",
   "tg_tps": 31.07,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 129/422",
   "hardware": "nvidia-tesla-p100",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "29b973d5e3e7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/777",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 361.47,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.68,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 130/422",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1f18dc387c45",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/561",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 359.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 131/422",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e86c3b770f09",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/605",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6700 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 596.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.37,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 132/422",
   "hardware": "amd-radeon-rx-6700-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "408180396c39",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/369",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 349.41,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 133/422",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1e010e31c59c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1469",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 377.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.45,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 134/422",
   "hardware": "apple-m3-max-40c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a78b448d0817",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/688",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+24GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 306.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 135/422",
   "hardware": "apple-m1-max-24c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "dc567444f089",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/550",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 302.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 30.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 136/422",
   "hardware": "apple-m4-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "52e1c7acd0bb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/380",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 292.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 29.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 137/422",
   "hardware": "apple-m4-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2d530a0be9db",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/470",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 8P+4E+19GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 261.76,
   "pp_test": "LocalScore avg",
   "tg_tps": 26.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 138/422",
   "hardware": "apple-m2-pro-19c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "800c81f2a980",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/428",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 283.81,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 139/422",
   "hardware": "apple-m3-pro-18c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1ed50dfa16fa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4201",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M6 2P+4E+12GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 263.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.24,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 140/422",
   "hardware": "apple-m6-12c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5da6d8b15fc1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/401",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 279.21,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.81,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 141/422",
   "hardware": "apple-m3-pro-18c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "db83150c5143",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/804",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 5700 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 256.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 142/422",
   "hardware": "amd-radeon-rx-5700-xt",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "68e4618ef351",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1078",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6600",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 418.8,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.24,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 143/422",
   "hardware": "amd-radeon-rx-6600",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3759be715dc4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1220",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 6P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 224.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 144/422",
   "hardware": "apple-m2-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cb5fe0aed6b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1298",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 6P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 216.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 148/422",
   "hardware": "apple-m2-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "45513be1c017",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/862",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 223.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 149/422",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d6cdaad48015",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/263",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 217.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 150/422",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bb630c29c60f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/929",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 204.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 151/422",
   "hardware": "apple-m1-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c44f37f646d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2912",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 202.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.62,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 152/422",
   "hardware": "apple-m5-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5abfd1eb2fe2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 200.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 153/422",
   "hardware": "apple-m1-pro-16c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "15551bced298",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1716",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 188.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.24,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 156/422",
   "hardware": "apple-m5-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a10d97a3cc34",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3811",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 176.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 158/422",
   "hardware": "apple-m5-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "eebf6d53b163",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/412",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 176.97,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 159/422",
   "hardware": "apple-m1-pro-14c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e5987dfddb44",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/418",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 173.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 160/422",
   "hardware": "apple-m1-pro-14c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "48b1c9e29435",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1556",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7600",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 317.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 162/422",
   "hardware": "amd-radeon-rx-7600",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c83a77aa3094",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/847",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 168.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.43,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 163/422",
   "hardware": "apple-m2-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b2383a8ee427",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/32",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 175.22,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 164/422",
   "hardware": "apple-m4-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f7c590d04a0e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/546",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 166.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.81,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 166/422",
   "hardware": "apple-m4-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1a7765543ef1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3710",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon™ RX 7700S",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 289.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 167/422",
   "hardware": "amd-radeon-rx-7700s",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c4a4ee940966",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/245",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 143.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 17.32,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 168/422",
   "hardware": "apple-m2-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a3a2c79dc86c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1085",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 147.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 169/422",
   "hardware": "apple-m4-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b533cd218a72",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/202",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 142.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.74,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 170/422",
   "hardware": "apple-m2-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "52290894ff31",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/499",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 141.26,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.65,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 171/422",
   "hardware": "apple-m2-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2dbb8f8465ca",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1394",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 134.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.27,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 172/422",
   "hardware": "apple-m4-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7c375b0d5cf2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/414",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 138.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 173/422",
   "hardware": "apple-m3-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2ccee79f4c90",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/17",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 134.01,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 174/422",
   "hardware": "apple-m1-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "57e3644118af",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1642",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 279.67,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.75,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 178/422; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-3050-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "1dbcdc4c97ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3123",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 124.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 179/422",
   "hardware": "apple-m3-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d1948f34ff5c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3128",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple A18 Pro 2P+4E+5GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 115.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.83,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 180/422",
   "hardware": "apple-a18-pro-5-core-gpu",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "57ae54787aa7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3235",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 186.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.63,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 182/422",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "90df0293ab69",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/366",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 290.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 184/422; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-3050-ti-laptop",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "a912ae8d6cd7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1939",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 113.96,
   "pp_test": "LocalScore avg",
   "tg_tps": 12.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 186/422",
   "hardware": "apple-m2-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "40ab9d7e8cb3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1281",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+7GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 108.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 187/422",
   "hardware": "apple-m1-7c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f0d9d948883d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1595",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1660 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 129.2,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 190/422",
   "hardware": "nvidia-gtx-1660-super",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "134c2eab3b60",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/727",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1660",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 128.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 192/422",
   "hardware": "nvidia-gtx-1660",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d5343d752157",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/562",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 109.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 193/422",
   "hardware": "apple-m3-10c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "545ce8517aec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/432",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 92.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 205/422",
   "hardware": "apple-m1-8c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "47e05851a9d5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/223",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+7GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 83.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 211/422",
   "hardware": "apple-m1-7c",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d77dff36e08b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1578",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon(TM) 780M",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 111.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 226/422",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "676dd544d149",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3179",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 109.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 236/422",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "accdeb8ff75d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3862",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 103.43,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 241/422",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2d3e6b40dba0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1965",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 93.86,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 251/422",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "96af51078c9f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4232",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA T1000 8GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 74.5,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 254/422",
   "hardware": "nvidia-t1000-8gb",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5b292cfc7a6b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/374",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1650 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 65.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 2.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 288/422; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-gtx-1650-ti",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "8e3344ef3c16",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3526",
   "model_page": "https://www.localscore.ai/model/1",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P620",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 45.39,
   "pp_test": "LocalScore avg",
   "tg_tps": 1.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 357/422; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-quadro-p620",
   "model": "llama-3-1-8b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "65e82f987595",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/937",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4416.36,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.05,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 2/418",
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "db3ce78bc492",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1704",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4208.36,
   "pp_test": "LocalScore avg",
   "tg_tps": 68.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 3/418",
   "hardware": "nvidia-rtx-4090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "278c58b9b842",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/970",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090 D",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3905.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 70.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 4/418",
   "hardware": "nvidia-rtx-4090-d",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3f4a6f06e2c6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/673",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3552.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 70.04,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 5/418",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3fd5602b146b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/69",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2973.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 72.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 6/418",
   "hardware": "nvidia-h100-pcie",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1d98c5d6e6bb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/77",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3506.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 49.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 7/418",
   "hardware": "nvidia-rtx-4090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "246ca0a16f73",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/155",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3678.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 8/418",
   "hardware": "nvidia-rtx-5090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "21b8fe026120",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1125",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3789.6,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.95,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 9/418",
   "hardware": "nvidia-rtx-5090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "58aeaa1492d7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2949",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2802.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.15,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 10/418",
   "hardware": "nvidia-rtx-4080-super",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b10e974662f2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/127",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2202.06,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.18,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 11/418",
   "hardware": "nvidia-rtx-3090-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "706459440088",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/61",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100-SXM4-80GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2068.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 66.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 12/418",
   "hardware": "nvidia-a100-80gb-sxm",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fac018d6bd3a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/102",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2680.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 43.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 13/418",
   "hardware": "nvidia-rtx-4080-super",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "adfed889087a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1553",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3055.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 31.57,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 14/418",
   "hardware": "nvidia-rtx-4090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ea2aad399be2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/86",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2935.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 28.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 15/418",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "75f6cf198afd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1950.01,
   "pp_test": "LocalScore avg",
   "tg_tps": 56.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 16/418",
   "hardware": "nvidia-rtx-3090",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9c18e6035c53",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/50",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1882.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 51.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 17/418",
   "hardware": "nvidia-rtx-a6000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e74eaa0a6202",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/138",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2176.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.18,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 18/418",
   "hardware": "nvidia-rtx-4070-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "27b840f548c2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/15",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2339.45,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 19/418",
   "hardware": "nvidia-rtx-4080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ed4d59de83c5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1119",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2415.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 33.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 20/418",
   "hardware": "nvidia-rtx-5080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8715d831ccb9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/82",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2712.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 21/418",
   "hardware": "nvidia-l40",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f8372a6960f5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/136",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2146.84,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 22/418",
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "da6730b5e59b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/87",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 80GB HBM3",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2608.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 23/418",
   "hardware": "nvidia-h100-sxm",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d420a0a85694",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/160",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2192.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 24/418",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7a959045509e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1344",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2622.65,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 25/418",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "17c5fe4b97a3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/49",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1825.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 26/418",
   "hardware": "nvidia-a40",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2f510b9024fb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/120",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1792.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 27/418",
   "hardware": "nvidia-rtx-3080-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c6168cb74b4e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2948",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1741.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 46.62,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 28/418",
   "hardware": "nvidia-rtx-3080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bfe9489a9272",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/84",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2867.59,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 29/418",
   "hardware": "nvidia-l40s",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cbb815405391",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/26",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1532.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 50.43,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 30/418",
   "hardware": "nvidia-rtx-a5000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3789c8cf8f89",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/78",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 NVL",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2353.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 31/418",
   "hardware": "nvidia-h100-nvl",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b0f07be91746",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3109",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A40-48Q",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1707.41,
   "pp_test": "LocalScore avg",
   "tg_tps": 39.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 32/418",
   "hardware": "nvidia-a40-48q",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f1c5fbd481b4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1524",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2580.79,
   "pp_test": "LocalScore avg",
   "tg_tps": 17.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 33/418",
   "hardware": "nvidia-rtx-pro-6000-blackwell-server-edition",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bc886c2e3f3a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/147",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1691.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 34/418",
   "hardware": "nvidia-rtx-4070",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "725a11fca925",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/489",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2097.1,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 35/418",
   "hardware": "nvidia-rtx-5080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ab6b43ce98fc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3553",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1549.65,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 36/418",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2d64f16b39fc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3684",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1757.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 37/418",
   "hardware": "nvidia-rtx-5070",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "003c61887d12",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/775",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1183.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 52.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 38/418",
   "hardware": "nvidia-tesla-v100-16gb",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a441b5c58dd3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1847",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1794.8,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 39/418",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "917a6e9fb6d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3548",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1473.59,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 40/418",
   "hardware": "nvidia-rtx-5070-ti-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e53893acdb04",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/51",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4500",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1330.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 43.98,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 41/418",
   "hardware": "nvidia-rtx-a4500",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ed52cba54222",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/920",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-32GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1178.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 49.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 42/418",
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4f83e2d5b002",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/315",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1701.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.88,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 43/418",
   "hardware": "nvidia-rtx-4070-super",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b0f49c73a8f1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3698",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 4500 Blackwell",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1824.39,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 44/418",
   "hardware": "nvidia-rtx-pro-4500-blackwell",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "99cc1e9ced0c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1399",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN RTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1199.97,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 45/418",
   "hardware": "nvidia-titan-rtx",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "416859b931df",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3817",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN RTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1229.41,
   "pp_test": "LocalScore avg",
   "tg_tps": 40.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 46/418",
   "hardware": "nvidia-titan-rtx",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b253221ac860",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/569",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1177.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 47/418",
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6b24bf7caf8a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/76",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1416.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 29.99,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 48/418",
   "hardware": "nvidia-rtx-4000-ada",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8b112d35a818",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/112",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1139.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 39.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 49/418",
   "hardware": "nvidia-rtx-2080-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "364f2d80598f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/860",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1329.1,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 50/418",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "412a7f001b2e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3048",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1086.86,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 51/418",
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8f8d4b28e8ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3523",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1076.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 52/418",
   "hardware": "nvidia-quadro-rtx-6000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e6f80fb95484",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/107",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN V",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 961.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 53/418",
   "hardware": "nvidia-titan-v",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2568072266ad",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/64",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100 80GB PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1437.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 54/418",
   "hardware": "nvidia-a100-80gb-pcie",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "412f19694240",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/632",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 3500 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1297.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 55/418",
   "hardware": "nvidia-rtx-3500-ada-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d118864fd592",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/7",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1238.62,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 56/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "6234e1851baa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2298",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1283.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.55,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 57/418",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cf54ac0c50c2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/132",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1254.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 58/418",
   "hardware": "nvidia-rtx-3080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5c96f14fe7b1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/44",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1041.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.03,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 59/418",
   "hardware": "nvidia-rtx-a4000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1c3e79d0c34e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/168",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1264.45,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 60/418",
   "hardware": "nvidia-rtx-5070",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "531ab2778fd8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3506",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-PCIE-32GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 906.96,
   "pp_test": "LocalScore avg",
   "tg_tps": 28.93,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 61/418",
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ef4fb7a4ec1c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/63",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L4",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1095.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 62/418",
   "hardware": "nvidia-l4",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4acc4aeef919",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/157",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 SFF Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1051.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 63/418",
   "hardware": "nvidia-rtx-4000-sff-ada",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7e035e2744c3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/363",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1133.16,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 64/418",
   "hardware": "nvidia-a10",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3d5fe7a56b5f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1771",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1118.67,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.37,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 65/418",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "617a79930dbb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/894",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 858.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 28.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 66/418",
   "hardware": "nvidia-rtx-3080-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0b6dc6b0806c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3816",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 2000 Blackwell",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 934.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 67/418",
   "hardware": "nvidia-rtx-pro-2000-blackwell",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fd41eebdd5de",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2944",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1068.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 68/418",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7abb59ed7ca6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1024",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 839.43,
   "pp_test": "LocalScore avg",
   "tg_tps": 27.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 69/418",
   "hardware": "nvidia-rtx-a5000-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bce0cb5d8e03",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/43",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 759.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 26.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 70/418",
   "hardware": "nvidia-rtx-3060",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "99719a9c3e5c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1357",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 5000 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 951.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 71/418",
   "hardware": "nvidia-rtx-5000-ada-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "961383b9fe05",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/404",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 580.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 72/418",
   "hardware": "apple-m3-ultra-80c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ef97c0e0f711",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1359",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 568.25,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 73/418",
   "hardware": "apple-m3-ultra-80c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "61b865e94ff0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/62",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 2000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 740.37,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.67,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 74/418",
   "hardware": "nvidia-rtx-2000-ada",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1fa3540f9af9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1222",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+76GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 470.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.63,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 75/418",
   "hardware": "apple-m2-ultra-76c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "efa73ff6a8cf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3670",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+76GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 469.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 76/418",
   "hardware": "apple-m2-ultra-76c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9a61228dbc73",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1967",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 594.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.82,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 77/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-2060",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "91ae73d95761",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4247",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 424.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 42.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 78/418",
   "hardware": "apple-m5-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "452b02bdf2fc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/479",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 20P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 444.62,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 79/418",
   "hardware": "apple-m3-ultra-60c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ee6c71b90ec8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/880",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 501.86,
   "pp_test": "LocalScore avg",
   "tg_tps": 27.95,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 80/418",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "001989117014",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2309",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 381.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 81/418",
   "hardware": "apple-m2-ultra-60c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2bfb095f1053",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/617",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 380.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 33.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 82/418",
   "hardware": "apple-m2-ultra-60c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "95d33f96c7d1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/491",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 479.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 83/418",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a3022f2b35ec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3148",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 359.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 36.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 84/418",
   "hardware": "apple-m5-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "dc91b08e3085",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/514",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 12GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 506.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 85/418",
   "hardware": "nvidia-rtx-a2000-12gb",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8047189bae99",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/475",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 512.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 86/418",
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "01679e6aa493",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1087",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+64GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 371.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 32.42,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 87/418",
   "hardware": "apple-m1-ultra-64c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0a352060f038",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/319",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6900 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 568.81,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 88/418",
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "25c8b03576fb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3105",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 343.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.15,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 89/418",
   "hardware": "apple-m5-max-32c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1110564006d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/108",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN Xp",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 411.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 90/418",
   "hardware": "nvidia-titan-xp",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "90ae032fa641",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1188",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 335.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 30.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 91/418",
   "hardware": "apple-m4-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "97509e6e1f39",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/103",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 418.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 92/418",
   "hardware": "nvidia-gtx-1080-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8332b740ceec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/6",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 328.43,
   "pp_test": "LocalScore avg",
   "tg_tps": 28.95,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 93/418",
   "hardware": "apple-m4-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7d3d25e9187d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/769",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla T4",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 449.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.32,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 95/418",
   "hardware": "nvidia-tesla-t4",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ddd4a2f93c26",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/306",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+48GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 289.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 27.77,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 96/418",
   "hardware": "apple-m1-ultra-48c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bd6a0269c945",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/750",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 302.37,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.49,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 97/418",
   "hardware": "apple-m3-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "20b415f83fa9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/776",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 287.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.62,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 98/418",
   "hardware": "apple-m4-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8a43463ccda7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/162",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 10P+4E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 284.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 99/418",
   "hardware": "apple-m4-max-32c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "15b03791d40e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1003",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD  Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 302.02,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.99,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 100/418",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "19ac17c37c1f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3986",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6750 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 351.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 101/418",
   "hardware": "amd-radeon-rx-6750-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "db2d77b6c587",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1114",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 286.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 102/418",
   "hardware": "apple-m3-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f1ec33494f87",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/494",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 399.2,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 103/418",
   "hardware": "amd-radeon-rx-6800",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "42318a3e7ae0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1729",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 329.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 104/418",
   "hardware": "nvidia-tesla-p40",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e9c993f8eb0a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/515",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 252.47,
   "pp_test": "LocalScore avg",
   "tg_tps": 25.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 105/418",
   "hardware": "apple-m2-max-38c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ee2a33237fb0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/607",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 272.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 106/418",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "49547cde8d3f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1106",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P5000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 302.68,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.86,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 107/418",
   "hardware": "nvidia-quadro-p5000",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b16a407edb52",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/605",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6700 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 316.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 108/418",
   "hardware": "amd-radeon-rx-6700-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2834b64bf48a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/789",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 238.76,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.77,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 109/418",
   "hardware": "apple-m3-max-30c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8fd0070e4320",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1291",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 225.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 110/418",
   "hardware": "apple-m2-max-38c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b8d259137a2a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3180",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 216.88,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 111/418",
   "hardware": "apple-m5-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0de04b3ca260",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3226",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 218.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.0,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 112/418",
   "hardware": "apple-m5-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "83243738ef1f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1577",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 222.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.96,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 113/418",
   "hardware": "apple-m2-max-38c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5f6a8ab867c5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3461",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 215.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 23.73,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 114/418",
   "hardware": "apple-m5-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8b43cc2bbf43",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/100",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 204.1,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 115/418",
   "hardware": "nvidia-tesla-p100",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ba6f68906ae5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1469",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 232.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 17.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 116/418",
   "hardware": "apple-m3-max-40c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1c797529a303",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/805",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 215.74,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.38,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 117/418",
   "hardware": "apple-m3-max-30c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e1536d6436e6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1664",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6950 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 549.35,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 119/418",
   "hardware": "amd-radeon-rx-6950-xt",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "939955a8bc89",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/597",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 196.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 120/418",
   "hardware": "apple-m1-max-32c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a20142138ac4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/364",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 191.39,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 121/418",
   "hardware": "apple-m1-max-32c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "aa52a874b3c2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/121",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 429.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 122/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-3060-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "e5c5595856cb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3323",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 5P+10E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 176.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 123/418",
   "hardware": "apple-m5-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b4c778fdfe29",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3387",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-PCIE-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 180.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.15,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 124/418",
   "hardware": "nvidia-tesla-p100",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "68b55d0b828c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/561",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 190.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 17.98,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 125/418",
   "hardware": "apple-m4-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "557d631db881",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1268",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4500 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 302.25,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 126/418",
   "hardware": "nvidia-rtx-a4500-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5fbcf0ae06e8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3856",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon AI PRO R9700",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 381.14,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 127/418",
   "hardware": "amd-radeon-ai-pro-r9700",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "416ff981f8ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3531",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 5P+10E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 169.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 128/418",
   "hardware": "apple-m5-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "36dc41097cc9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/777",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 186.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 17.57,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 129/418",
   "hardware": "apple-m4-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c90fa1bb785a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/369",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 182.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 130/418",
   "hardware": "apple-m4-pro-20c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9f1d09144dda",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1943",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+24GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 154.75,
   "pp_test": "LocalScore avg",
   "tg_tps": 18.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 132/418",
   "hardware": "apple-m1-max-24c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0bd827ef06f1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/126",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 398.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.62,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 133/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-3070",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "a52344e8c68b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/550",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 156.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 136/418",
   "hardware": "apple-m4-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "06605ddbbd95",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/380",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 151.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 137/418",
   "hardware": "apple-m4-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b48ce829e074",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2277",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 151.01,
   "pp_test": "LocalScore avg",
   "tg_tps": 16.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 138/418",
   "hardware": "apple-m4-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3de3396e7adf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1473",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 149.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.49,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 140/418",
   "hardware": "apple-m2-max-30c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8e67a396b236",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/688",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+24GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 140.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 15.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 141/418",
   "hardware": "apple-m1-max-24c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "43c33273d0ec",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1476",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 323.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.32,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 142/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-2080",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "c1e7daa381e5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/470",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 8P+4E+19GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 137.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 14.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 143/418",
   "hardware": "apple-m2-pro-19c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c0a8157f831d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1797",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S-8Q",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 343.21,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.45,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 144/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-l40s-8q",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "cdf6cf83923b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/428",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 147.29,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 145/418",
   "hardware": "apple-m3-pro-18c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cd2c7aa5fe4c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/401",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 145.35,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 146/418",
   "hardware": "apple-m3-pro-18c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "43b8a53141f0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2011",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 293.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 147/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-2080-super",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "7ff299fb542b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1220",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 6P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 118.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 148/418",
   "hardware": "apple-m2-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3731b611c27a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/328",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 326.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 149/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "479a68960b6e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/862",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 119.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 12.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 150/418",
   "hardware": "apple-m3-pro-14c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "da64fb6bd935",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1537",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 288.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 153/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4070-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "65a6605db630",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/263",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 114.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.61,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 154/418",
   "hardware": "apple-m3-pro-14c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "827d70d52106",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2912",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 110.42,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 157/418",
   "hardware": "apple-m5-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9bf4b168af6d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/929",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 106.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 158/418",
   "hardware": "apple-m1-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "51c9b057524e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 103.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 160/418",
   "hardware": "apple-m1-pro-16c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d4ab2c6df15a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3738",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A2",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 159.75,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 162/418",
   "hardware": "nvidia-a2",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b5ab564847fb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/412",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 92.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.83,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 167/418",
   "hardware": "apple-m1-pro-14c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "159ea3dd3985",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1602",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 88.87,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 170/418",
   "hardware": "apple-m1-pro-14c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "95b0e4d20d02",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1085",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 92.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.08,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 171/418",
   "hardware": "apple-m4-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "baa37ba52428",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/32",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 87.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 8.99,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 173/418",
   "hardware": "apple-m4-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e4a18f0552d9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3811",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 81.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 8.85,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 175/418",
   "hardware": "apple-m5-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "170594b1a013",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/418",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 82.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 8.93,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 176/418",
   "hardware": "apple-m1-pro-14c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "495c04a0b3cf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/546",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 79.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 8.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 178/418",
   "hardware": "apple-m4-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "999a85efe4c5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1487",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 2000 Blackwell Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 229.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 2.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 181/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-pro-2000-blackwell-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "95a6da846fae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/499",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 72.43,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 182/418",
   "hardware": "apple-m2-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "06067eb279b6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/202",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 70.26,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 185/418",
   "hardware": "apple-m2-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fcf39321ad4b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3235",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 99.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 3.05,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 191/418",
   "hardware": "amd-radeon-780m",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "903fd1af3a84",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1394",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 60.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.26,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 193/418",
   "hardware": "apple-m4-8c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0f7c6c6e802e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1285",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Instinct MI50/MI60",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 48.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 12.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 194/418",
   "hardware": "amd-instinct-mi50",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "167f9c7c2472",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/562",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 65.8,
   "pp_test": "LocalScore avg",
   "tg_tps": 6.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 195/418",
   "hardware": "apple-m3-10c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e0c68053dccc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1939",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 59.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 6.96,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 197/418",
   "hardware": "apple-m2-8c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0e7f531763df",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/432",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 53.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.43,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 209/418",
   "hardware": "apple-m1-8c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cf879d489f17",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/223",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+7GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 40.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 4.81,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 241/418",
   "hardware": "apple-m1-7c",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fe3b426d4c89",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2943",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 57.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 1.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 258/418",
   "hardware": "amd-radeon-780m",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e91c41d7857b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/767",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 82.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 1.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 269/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-3060-laptop",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "f36b4758c766",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1009",
   "model_page": "https://www.localscore.ai/model/2",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1050 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 36.86,
   "pp_test": "LocalScore avg",
   "tg_tps": 1.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 326/418; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-gtx-1050-ti",
   "model": "qwen2-5-14b-instruct",
   "partial_offload": true,
   "method": "LocalScore",
   "id": "3660a5c51334",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/673",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 27281.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 410.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 1/878",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "eb2f6e5d52c3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1704",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 25538.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 381.74,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 2/878",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7e3f7cd03346",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/970",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090 D",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 24548.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 359.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 3/878",
   "hardware": "nvidia-rtx-4090-d",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c82ae8b70dbe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1653",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 24809.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 354.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 4/878",
   "hardware": "nvidia-l40s",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ff967c4bed73",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/69",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 18590.42,
   "pp_test": "LocalScore avg",
   "tg_tps": 335.43,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 5/878",
   "hardware": "nvidia-h100-pcie",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6f8ec492655b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/127",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 15119.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 353.86,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 6/878",
   "hardware": "nvidia-rtx-3090-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c08f9f48b7d5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/61",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100-SXM4-80GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 15370.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 308.11,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 7/878",
   "hardware": "nvidia-a100-80gb-sxm",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c22930f37c07",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2949",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 17478.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 247.22,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 8/878",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e4f920e405bb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/77",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 19320.79,
   "pp_test": "LocalScore avg",
   "tg_tps": 202.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 9/878",
   "hardware": "nvidia-rtx-4090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0c780cf8c274",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/120",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 14142.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 312.81,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 10/878",
   "hardware": "nvidia-rtx-3080-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c60cb2e78603",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/50",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13191.25,
   "pp_test": "LocalScore avg",
   "tg_tps": 315.3,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 11/878",
   "hardware": "nvidia-rtx-a6000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1040dd2445dd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/86",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 6000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 19619.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 130.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 12/878",
   "hardware": "nvidia-rtx-6000-ada",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b5c0691e15c6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13482.32,
   "pp_test": "LocalScore avg",
   "tg_tps": 269.93,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 13/878",
   "hardware": "nvidia-rtx-3090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4301b03ab6d3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/155",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 20305.31,
   "pp_test": "LocalScore avg",
   "tg_tps": 170.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 14/878",
   "hardware": "nvidia-rtx-5090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f6ced6d3750e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/26",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12137.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 310.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 15/878",
   "hardware": "nvidia-rtx-a5000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4e7cc3f65b2e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/937",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 20439.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 191.37,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 16/878",
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c9101e0773fb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/147",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12717.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 282.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 17/878",
   "hardware": "nvidia-rtx-4070",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "eaea551db74f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4068",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12200.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 287.85,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 18/878",
   "hardware": "nvidia-rtx-a6000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "359583f4cf31",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/15",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 14496.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 241.03,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 19/878",
   "hardware": "nvidia-rtx-4080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "931b957f6098",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/49",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12156.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 283.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 20/878",
   "hardware": "nvidia-a40",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0bbba033dbfc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/84",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 18254.98,
   "pp_test": "LocalScore avg",
   "tg_tps": 120.98,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 21/878",
   "hardware": "nvidia-l40s",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6a33b17b7cd1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/102",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 16736.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 140.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 22/878",
   "hardware": "nvidia-rtx-4080-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "31d02626ff12",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/82",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 17685.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 118.02,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 23/878",
   "hardware": "nvidia-l40",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1a753e95c9c7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2948",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11748.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 254.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 24/878",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ac037004b35e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1524",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Server Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 18081.14,
   "pp_test": "LocalScore avg",
   "tg_tps": 105.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 25/878",
   "hardware": "nvidia-rtx-pro-6000-blackwell-server-edition",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ee9fd4eb6d70",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/51",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4500",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 10421.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 286.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 26/878",
   "hardware": "nvidia-rtx-a4500",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "93397c2f5056",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/315",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13280.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 182.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 27/878",
   "hardware": "nvidia-rtx-4070-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f6f5f7a09ff4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/132",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11589.2,
   "pp_test": "LocalScore avg",
   "tg_tps": 228.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 28/878",
   "hardware": "nvidia-rtx-3080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2b398a26b04a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/122",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 10074.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 297.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 29/878",
   "hardware": "nvidia-rtx-3070-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d3310b79461e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1344",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 16183.58,
   "pp_test": "LocalScore avg",
   "tg_tps": 106.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 30/878",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9c55e31b0d24",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/775",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9853.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 267.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 31/878",
   "hardware": "nvidia-tesla-v100-16gb",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8882adbe6a47",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/136",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13267.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 160.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 32/878",
   "hardware": "nvidia-rtx-4070-ti-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "332bb7d2fcb4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/920",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-SXM2-32GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9832.85,
   "pp_test": "LocalScore avg",
   "tg_tps": 264.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 33/878",
   "hardware": "nvidia-tesla-v100-32gb",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "af7a531f8b2d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3548",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11611.97,
   "pp_test": "LocalScore avg",
   "tg_tps": 175.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 34/878",
   "hardware": "nvidia-rtx-5070-ti-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d6b70221e8c0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/138",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12817.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 165.45,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 35/878",
   "hardware": "nvidia-rtx-4070-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c2c4fb396cbe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3553",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11598.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 182.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 36/878",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a9a1645358e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2197",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11965.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 207.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 37/878",
   "hardware": "nvidia-rtx-4090-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "159e646d154a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/569",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9356.26,
   "pp_test": "LocalScore avg",
   "tg_tps": 250.36,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 38/878",
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3e4f81dd3f0a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/22",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100-PCIE-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8984.17,
   "pp_test": "LocalScore avg",
   "tg_tps": 248.26,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 39/878",
   "hardware": "nvidia-tesla-v100-16gb",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e1392c13fb66",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/94",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H200",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 14696.3,
   "pp_test": "LocalScore avg",
   "tg_tps": 98.49,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 40/878",
   "hardware": "nvidia-h200",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "856050f76a8f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3423",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9137.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 252.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 41/878",
   "hardware": "nvidia-quadro-rtx-6000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "247f5b1083dc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/87",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 80GB HBM3",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 14607.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 96.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 42/878",
   "hardware": "nvidia-h100-sxm",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "77fd0ad1e813",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/76",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 10448.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 210.15,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 43/878",
   "hardware": "nvidia-rtx-4000-ada",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c8625cbe3bc1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/160",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 14814.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 197.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 44/878",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f8ddeef2b4bc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3817",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN RTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9273.14,
   "pp_test": "LocalScore avg",
   "tg_tps": 217.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 45/878",
   "hardware": "nvidia-titan-rtx",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e9d591063105",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/112",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8722.01,
   "pp_test": "LocalScore avg",
   "tg_tps": 222.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 46/878",
   "hardware": "nvidia-rtx-2080-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3cf74847ff8d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/78",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA H100 NVL",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13586.41,
   "pp_test": "LocalScore avg",
   "tg_tps": 83.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 47/878",
   "hardware": "nvidia-h100-nvl",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8ca91af20595",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/7",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9225.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 210.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 48/878",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ff8275c71ed3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/107",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN V",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7852.81,
   "pp_test": "LocalScore avg",
   "tg_tps": 231.55,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 49/878",
   "hardware": "nvidia-titan-v",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "703d3fc2aaf1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3048",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 8000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8204.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 218.08,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 50/878",
   "hardware": "nvidia-quadro-rtx-8000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "04bcc1ce3ae2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/63",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L4",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9268.37,
   "pp_test": "LocalScore avg",
   "tg_tps": 187.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 51/878",
   "hardware": "nvidia-l4",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1e598af176d0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/126",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8074.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 214.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 52/878",
   "hardware": "nvidia-rtx-3070",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a6d0b0db6731",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/157",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 4000 SFF Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8753.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 189.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 53/878",
   "hardware": "nvidia-rtx-4000-sff-ada",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3c61f7c4bf0d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/632",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 3500 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8206.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 181.38,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 54/878",
   "hardware": "nvidia-rtx-3500-ada-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7c3ebe1c85d4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4130",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro RTX 5000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7219.77,
   "pp_test": "LocalScore avg",
   "tg_tps": 227.89,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 55/878",
   "hardware": "nvidia-quadro-rtx-5000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "af802e1a794c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/121",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7643.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 192.96,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 56/878",
   "hardware": "nvidia-rtx-3060-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f8c5c923fbfd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/64",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A100 80GB PCIe",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 10805.83,
   "pp_test": "LocalScore avg",
   "tg_tps": 81.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 57/878",
   "hardware": "nvidia-a100-80gb-pcie",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e5c96e4bcd30",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/44",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7751.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 179.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 58/878",
   "hardware": "nvidia-rtx-a4000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8204f565f42c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1797",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA L40S-8Q",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 12563.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 62.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 59/878",
   "hardware": "nvidia-l40s-8q",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b765d688ddfe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1476",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6746.79,
   "pp_test": "LocalScore avg",
   "tg_tps": 210.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 60/878",
   "hardware": "nvidia-rtx-2080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "efb5ccee4706",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4016",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 11677.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 138.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 61/878",
   "hardware": "nvidia-rtx-5090-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f4d5f2199c64",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2011",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2080 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7096.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 186.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 62/878",
   "hardware": "nvidia-rtx-2080-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d79b3735d2bf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1024",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A5000 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7018.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 201.95,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 63/878",
   "hardware": "nvidia-rtx-a5000-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c49c8ac32629",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/860",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9434.68,
   "pp_test": "LocalScore avg",
   "tg_tps": 214.11,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 64/878",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d749ad6ab606",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/489",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 15538.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 196.79,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 65/878",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0c89ff82fd99",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1847",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 9863.08,
   "pp_test": "LocalScore avg",
   "tg_tps": 92.78,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 66/878",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "713674b6e48d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/894",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3080 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7533.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 157.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 67/878",
   "hardware": "nvidia-rtx-3080-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e4c20523c6a6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1119",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 10982.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 83.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 68/878",
   "hardware": "nvidia-rtx-5080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ce4865037049",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2298",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8192.88,
   "pp_test": "LocalScore avg",
   "tg_tps": 125.93,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 69/878",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "dfa9e28dc130",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1125",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 13214.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 69.32,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 70/878",
   "hardware": "nvidia-rtx-5090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bb5cfd9279b1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2944",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7523.85,
   "pp_test": "LocalScore avg",
   "tg_tps": 148.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 71/878",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5ba780d4e502",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/363",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA A10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8777.06,
   "pp_test": "LocalScore avg",
   "tg_tps": 93.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 72/878",
   "hardware": "nvidia-a10",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1de3fecadc4c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1537",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7398.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 142.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 73/878",
   "hardware": "nvidia-rtx-4070-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5d381415cdf5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/328",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7937.05,
   "pp_test": "LocalScore avg",
   "tg_tps": 123.0,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 74/878",
   "hardware": "nvidia-rtx-4060-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bf0f0d37deac",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/404",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5601.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 178.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 75/878",
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e6cd30c02af0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1727",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5815.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 190.19,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 76/878",
   "hardware": "nvidia-rtx-2070",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d4f5d4c5e800",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/43",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6042.27,
   "pp_test": "LocalScore avg",
   "tg_tps": 184.98,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 77/878",
   "hardware": "nvidia-rtx-3060",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "35fa03298fb7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3478",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2070 Super",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5696.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 189.11,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 78/878",
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "24c1cea6e3d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/111",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2060 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5615.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 199.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 79/878",
   "hardware": "nvidia-rtx-2060-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7688cdc08738",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3148",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4331.35,
   "pp_test": "LocalScore avg",
   "tg_tps": 262.67,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 80/878",
   "hardware": "apple-m5-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7d526679b693",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4247",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4283.25,
   "pp_test": "LocalScore avg",
   "tg_tps": 260.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 81/878",
   "hardware": "apple-m5-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "97db70f8b689",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/62",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 2000 Ada Generation",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6087.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 168.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 82/878",
   "hardware": "nvidia-rtx-2000-ada",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "23b428b77e91",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1359",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 24P+8E+80GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4999.5,
   "pp_test": "LocalScore avg",
   "tg_tps": 177.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 83/878",
   "hardware": "apple-m3-ultra-80c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "665d3840e592",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/529",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6339.75,
   "pp_test": "LocalScore avg",
   "tg_tps": 136.83,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 84/878",
   "hardware": "nvidia-rtx-4060-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c1eeb3218564",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1771",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 6729.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 116.07,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 85/878",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "be6e160d27d8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1487",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 2000 Blackwell Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5884.34,
   "pp_test": "LocalScore avg",
   "tg_tps": 130.96,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 86/878",
   "hardware": "nvidia-rtx-pro-2000-blackwell-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f5248dac1afe",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/767",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3060 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5429.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 151.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 87/878",
   "hardware": "nvidia-rtx-3060-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ee8ea5fa3e67",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/101",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2070 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5376.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 150.63,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 88/878",
   "hardware": "nvidia-rtx-2070-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a7445fe7305b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/479",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 20P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4393.74,
   "pp_test": "LocalScore avg",
   "tg_tps": 172.92,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 89/878",
   "hardware": "apple-m3-ultra-60c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ada4045f36f9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3105",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Max 6P+12E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3634.84,
   "pp_test": "LocalScore avg",
   "tg_tps": 229.29,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 90/878",
   "hardware": "apple-m5-max-32c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "367560a723dc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3572",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3070 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4930.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 124.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 91/878",
   "hardware": "nvidia-rtx-3070-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a7d6f0b261c1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/137",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5352.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 133.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 92/878",
   "hardware": "nvidia-rtx-4060",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cf991b23a95c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/6",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3848.14,
   "pp_test": "LocalScore avg",
   "tg_tps": 183.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 93/878",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "292cc8964310",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/168",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 8328.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 100.65,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 94/878",
   "hardware": "nvidia-rtx-5070",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "59ecd7be4958",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/776",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3816.8,
   "pp_test": "LocalScore avg",
   "tg_tps": 180.04,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 95/878",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1dfa42d9d6cf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/108",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA TITAN Xp",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4221.21,
   "pp_test": "LocalScore avg",
   "tg_tps": 160.38,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 96/878",
   "hardware": "nvidia-titan-xp",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ac3b40fb1860",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1188",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3750.57,
   "pp_test": "LocalScore avg",
   "tg_tps": 178.99,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 97/878",
   "hardware": "apple-m4-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "be9af7cce69b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/514",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 12GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4548.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 139.67,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 98/878",
   "hardware": "nvidia-rtx-a2000-12gb",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "04379878491e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/665",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4479.26,
   "pp_test": "LocalScore avg",
   "tg_tps": 136.55,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 99/878",
   "hardware": "nvidia-rtx-2060",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3e26635504ef",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/598",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4050 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4917.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 118.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 100/878",
   "hardware": "nvidia-rtx-4050-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "080a63377a1e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/103",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1080 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4098.22,
   "pp_test": "LocalScore avg",
   "tg_tps": 146.78,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 101/878",
   "hardware": "nvidia-gtx-1080-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "af0ca78f97d2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/617",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3295.52,
   "pp_test": "LocalScore avg",
   "tg_tps": 176.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 102/878",
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a89cb2e9a7ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2309",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3290.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 174.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 103/878",
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "91a086d007b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/273",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 16P+8E+60GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3272.21,
   "pp_test": "LocalScore avg",
   "tg_tps": 169.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 104/878",
   "hardware": "apple-m2-ultra-60c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "11441c2491cd",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/366",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 Ti Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4086.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 129.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 105/878",
   "hardware": "nvidia-rtx-3050-ti-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "417e73698840",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/614",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Radeon RX 7900 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4957.02,
   "pp_test": "LocalScore avg",
   "tg_tps": 99.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 106/878",
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c6428fc47ae2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/162",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max 10P+4E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3255.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 166.42,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 107/878",
   "hardware": "apple-m4-max-32c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d5ccf7327688",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1114",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3382.1,
   "pp_test": "LocalScore avg",
   "tg_tps": 150.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 108/878",
   "hardware": "apple-m3-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ddd280fe7ac7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1209",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon PRO V620",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4607.88,
   "pp_test": "LocalScore avg",
   "tg_tps": 101.79,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 109/878",
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f811aae729eb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/750",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3291.91,
   "pp_test": "LocalScore avg",
   "tg_tps": 146.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 110/878",
   "hardware": "apple-m3-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c305f82366ee",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1642",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3705.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 136.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 111/878",
   "hardware": "nvidia-rtx-3050-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6448f7b65dcf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1425",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3811.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 116.51,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 112/878",
   "hardware": "nvidia-rtx-a2000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4bbc866fecb8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1087",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+64GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2977.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 151.07,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 113/878",
   "hardware": "apple-m1-ultra-64c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c714ef02e464",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4147",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7900 GRE",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3609.88,
   "pp_test": "LocalScore avg",
   "tg_tps": 124.8,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 114/878",
   "hardware": "amd-radeon-rx-7900-gre",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "59cfac1372c8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1404",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A1000 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3664.83,
   "pp_test": "LocalScore avg",
   "tg_tps": 123.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 115/878",
   "hardware": "nvidia-rtx-a1000-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0d6fdc884469",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/475",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4376.52,
   "pp_test": "LocalScore avg",
   "tg_tps": 91.6,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 116/878",
   "hardware": "amd-radeon-rx-6800-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "41eee6c90736",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/585",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 9070 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2707.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 162.5,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 117/878",
   "hardware": "amd-radeon-rx-9070-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "40130df28793",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3653",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 7818.02,
   "pp_test": "LocalScore avg",
   "tg_tps": 184.75,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 118/878",
   "hardware": "nvidia-rtx-5060",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "bf4a3d262725",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/731",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 8GB Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3398.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 108.9,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 119/878",
   "hardware": "nvidia-rtx-a2000-8gb-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e980967b1dc2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1268",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A4500 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3868.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 72.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 120/878",
   "hardware": "nvidia-rtx-a4500-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "188f82cbf6b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3558",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 5678.48,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.72,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 121/878",
   "hardware": "nvidia-rtx-5080-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "be95f43a1665",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3180",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2412.83,
   "pp_test": "LocalScore avg",
   "tg_tps": 175.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 122/878",
   "hardware": "apple-m5-pro-20c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b2b0558429dc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1291",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2522.84,
   "pp_test": "LocalScore avg",
   "tg_tps": 150.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 123/878",
   "hardware": "apple-m2-max-38c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "779809d44043",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3226",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 Pro 6P+12E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2404.39,
   "pp_test": "LocalScore avg",
   "tg_tps": 174.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 124/878",
   "hardware": "apple-m5-pro-20c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "914055702f76",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/306",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Ultra 16P+4E+48GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2582.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 137.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 125/878",
   "hardware": "apple-m1-ultra-48c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "05cbf8062ea4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/977",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3347.07,
   "pp_test": "LocalScore avg",
   "tg_tps": 103.58,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 126/878",
   "hardware": "nvidia-rtx-3050",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "31ad1a569724",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/491",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4136.54,
   "pp_test": "LocalScore avg",
   "tg_tps": 82.04,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 127/878",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "90a7820e7568",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/25",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1070 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2980.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 115.77,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 128/878",
   "hardware": "nvidia-gtx-1070-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d52b4f738134",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/975",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX A2000 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3084.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 105.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 129/878",
   "hardware": "nvidia-rtx-a2000-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3352b4fde26d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1998",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3050 6GB Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3329.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 99.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 130/878",
   "hardware": "nvidia-rtx-3050-6gb-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "433841cf7158",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1469",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 12P+4E+40GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2750.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 120.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 131/878",
   "hardware": "apple-m3-max-40c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ea89200ba8f8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/805",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2553.3,
   "pp_test": "LocalScore avg",
   "tg_tps": 132.99,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 132/878",
   "hardware": "apple-m3-max-30c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4468f1487188",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/880",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Radeon RX 7900 XTX",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4041.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 74.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 133/878",
   "hardware": "amd-radeon-rx-7900-xtx",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "514513959105",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1225",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX 500 Ada Generation Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3624.19,
   "pp_test": "LocalScore avg",
   "tg_tps": 81.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 134/878",
   "hardware": "nvidia-rtx-500-ada-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c610f1c4fece",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1577",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+38GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2409.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 140.57,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 135/878",
   "hardware": "apple-m2-max-38c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b01ef4b75ffb",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/789",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Max 10P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2519.96,
   "pp_test": "LocalScore avg",
   "tg_tps": 132.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 136/878",
   "hardware": "apple-m3-max-30c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4adf92034bb5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/100",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-SXM2-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2515.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 124.25,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 137/878",
   "hardware": "nvidia-tesla-p100",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "500d475167a5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1003",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD  Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2512.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 120.68,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 138/878",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d391857b508a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/319",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6900 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 4411.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 62.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 139/878",
   "hardware": "amd-radeon-rx-6900-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "02e51285aa56",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1729",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P40",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2662.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 104.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 140/878",
   "hardware": "nvidia-tesla-p40",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "67d6f32902ba",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4073",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P6000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2873.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 86.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 141/878",
   "hardware": "nvidia-quadro-p6000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "5f83ea729812",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/98",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1070",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2547.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 102.34,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 142/878",
   "hardware": "nvidia-gtx-1070",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3b169a20ca86",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/607",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7800 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2934.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 112.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 143/878",
   "hardware": "amd-radeon-rx-7800-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8d94b0292f4a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3387",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla P100-PCIE-16GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2243.67,
   "pp_test": "LocalScore avg",
   "tg_tps": 115.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 144/878",
   "hardware": "nvidia-tesla-p100",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7c87955643cc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1079",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Max 8P+4E+30GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2031.67,
   "pp_test": "LocalScore avg",
   "tg_tps": 127.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 145/878",
   "hardware": "apple-m2-max-30c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "44c56e18c24e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/597",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2023.35,
   "pp_test": "LocalScore avg",
   "tg_tps": 125.59,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 146/878",
   "hardware": "apple-m1-max-32c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6cdb09a3492a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/110",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1080",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2784.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 80.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 147/878",
   "hardware": "nvidia-gtx-1080",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8eb04317a2df",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/561",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2130.82,
   "pp_test": "LocalScore avg",
   "tg_tps": 119.64,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 148/878",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e2355bb28dc6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/369",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2145.13,
   "pp_test": "LocalScore avg",
   "tg_tps": 118.56,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 149/878",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0a06ed97b73b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/494",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6800",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3322.28,
   "pp_test": "LocalScore avg",
   "tg_tps": 69.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 150/878",
   "hardware": "amd-radeon-rx-6800",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e1ce340eb97b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/12",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6650 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2641.14,
   "pp_test": "LocalScore avg",
   "tg_tps": 88.18,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 151/878",
   "hardware": "amd-radeon-rx-6650-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "37175b2bddf4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/364",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+32GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1941.65,
   "pp_test": "LocalScore avg",
   "tg_tps": 116.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 152/878",
   "hardware": "apple-m1-max-32c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0333ac40417d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/777",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 10P+4E+20GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2037.17,
   "pp_test": "LocalScore avg",
   "tg_tps": 112.04,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 153/878",
   "hardware": "apple-m4-pro-20c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b635af655618",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1740",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P4000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2194.22,
   "pp_test": "LocalScore avg",
   "tg_tps": 94.08,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 154/878",
   "hardware": "nvidia-quadro-p4000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "58fa3760de07",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2277",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1858.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 111.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 155/878",
   "hardware": "apple-m4-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c6669067270e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/380",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1806.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 110.84,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 156/878",
   "hardware": "apple-m4-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e962d5761851",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/605",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6700 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2883.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.27,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 157/878",
   "hardware": "amd-radeon-rx-6700-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e23cc519d184",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/688",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+24GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1669.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 105.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 159/878",
   "hardware": "apple-m1-max-24c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "82312bcbe661",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/550",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Pro 8P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1706.69,
   "pp_test": "LocalScore avg",
   "tg_tps": 105.47,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 160/878",
   "hardware": "apple-m4-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "963e3e935275",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1943",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max 8P+2E+24GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1582.03,
   "pp_test": "LocalScore avg",
   "tg_tps": 93.86,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 163/878",
   "hardware": "apple-m1-max-24c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6ef9850470a1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1289",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1060 6GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1862.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 76.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 165/878",
   "hardware": "nvidia-gtx-1060",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "1861681f7dea",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/470",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 8P+4E+19GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1487.56,
   "pp_test": "LocalScore avg",
   "tg_tps": 100.26,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 167/878",
   "hardware": "apple-m2-pro-19c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a32d245f4f4e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1862",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 8P+4E+19GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1457.87,
   "pp_test": "LocalScore avg",
   "tg_tps": 99.52,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 169/878",
   "hardware": "apple-m2-pro-19c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "0b1db0940f54",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/428",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1586.17,
   "pp_test": "LocalScore avg",
   "tg_tps": 89.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 170/878",
   "hardware": "apple-m3-pro-18c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4b50e7e798a3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/401",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 6P+6E+18GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1605.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 89.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 171/878",
   "hardware": "apple-m3-pro-18c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ee2cbe96956f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3134",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA P106-100",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1758.04,
   "pp_test": "LocalScore avg",
   "tg_tps": 70.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 172/878",
   "hardware": "nvidia-p106-100",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f20bf98ba7ff",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1220",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 6P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1327.11,
   "pp_test": "LocalScore avg",
   "tg_tps": 91.04,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 174/878",
   "hardware": "apple-m2-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f80eff4f69db",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3811",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1286.85,
   "pp_test": "LocalScore avg",
   "tg_tps": 99.79,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 176/878",
   "hardware": "apple-m5-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "56f6cdc5dd48",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1298",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Pro 6P+4E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1281.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 91.45,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 177/878",
   "hardware": "apple-m2-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8d35404c0807",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2912",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1275.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 98.18,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 178/878",
   "hardware": "apple-m5-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ef26e1b17893",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1716",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M5 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1253.4,
   "pp_test": "LocalScore avg",
   "tg_tps": 99.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 179/878",
   "hardware": "apple-m5-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "16fff38fed54",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/862",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1328.12,
   "pp_test": "LocalScore avg",
   "tg_tps": 87.33,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 181/878",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "86673312b219",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1078",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 6600",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 2028.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 52.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 183/878",
   "hardware": "amd-radeon-rx-6600",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6ca3552e6c08",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/263",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Pro 5P+6E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1293.5,
   "pp_test": "LocalScore avg",
   "tg_tps": 88.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 184/878",
   "hardware": "apple-m3-pro-14c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8cd113e0493b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1050",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7600 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1501.87,
   "pp_test": "LocalScore avg",
   "tg_tps": 57.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 186/878",
   "hardware": "amd-radeon-rx-7600-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f8dd06a1fab1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/929",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1158.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 78.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 188/878",
   "hardware": "apple-m1-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2cffc8105dde",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 8P+2E+16GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1170.51,
   "pp_test": "LocalScore avg",
   "tg_tps": 77.44,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 189/878",
   "hardware": "apple-m1-pro-16c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9767e60fbc58",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/736",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 2050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1769.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 47.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 190/878",
   "hardware": "nvidia-rtx-2050",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2d90747f569b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1624",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P2000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1353.42,
   "pp_test": "LocalScore avg",
   "tg_tps": 58.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 193/878",
   "hardware": "nvidia-quadro-p2000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e07846b9d9af",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/32",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1095.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 75.97,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 196/878",
   "hardware": "apple-m4-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ab5b49fd48e7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/418",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1062.36,
   "pp_test": "LocalScore avg",
   "tg_tps": 72.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 199/878",
   "hardware": "apple-m1-pro-14c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b376275158ed",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4063",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon Pro V620",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 3201.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 12.13,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 200/878",
   "hardware": "amd-radeon-pro-v620",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "7471d269103e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/412",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro 6P+2E+14GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1047.73,
   "pp_test": "LocalScore avg",
   "tg_tps": 72.1,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 201/878",
   "hardware": "apple-m1-pro-14c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8286db575ac6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/546",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1044.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 74.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 203/878",
   "hardware": "apple-m4-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "cb524fdcf1ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1085",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1027.65,
   "pp_test": "LocalScore avg",
   "tg_tps": 74.35,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 205/878",
   "hardware": "apple-m4-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fec240197ab6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/414",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 932.52,
   "pp_test": "LocalScore avg",
   "tg_tps": 67.12,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 209/878",
   "hardware": "apple-m3-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "158b0889f7ef",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/755",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Orin",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1157.2,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.26,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 210/878",
   "hardware": "nvidia-jetson-agx-orin-64gb",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "734fd47e8213",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/562",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 915.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 64.77,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 211/878",
   "hardware": "apple-m3-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2ef1b1991dff",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1394",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 4P+6E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 897.9,
   "pp_test": "LocalScore avg",
   "tg_tps": 65.93,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 212/878",
   "hardware": "apple-m4-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "45f2403e58e4",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1009",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1050 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 1013.64,
   "pp_test": "LocalScore avg",
   "tg_tps": 45.23,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 217/878",
   "hardware": "nvidia-gtx-1050-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "983ff853dc95",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/847",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 825.66,
   "pp_test": "LocalScore avg",
   "tg_tps": 58.69,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 223/878",
   "hardware": "apple-m2-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "68ea4308401c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3123",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 803.78,
   "pp_test": "LocalScore avg",
   "tg_tps": 60.31,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 224/878",
   "hardware": "apple-m3-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "2dc6ad337171",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/202",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 813.71,
   "pp_test": "LocalScore avg",
   "tg_tps": 54.76,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 225/878",
   "hardware": "apple-m2-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ca3e4a23f6d9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/499",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+10GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 801.84,
   "pp_test": "LocalScore avg",
   "tg_tps": 55.57,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 226/878",
   "hardware": "apple-m2-10c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "a1087d4cda28",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3136",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA P106-090",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 922.75,
   "pp_test": "LocalScore avg",
   "tg_tps": 39.22,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 227/878",
   "hardware": "nvidia-p106-090",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "010e54f493c9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/765",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1660 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 828.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.09,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 239/878",
   "hardware": "nvidia-gtx-1660-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e033c0491a2a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1595",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1660 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 798.61,
   "pp_test": "LocalScore avg",
   "tg_tps": 35.16,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 246/878",
   "hardware": "nvidia-gtx-1660-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "10c0a6fe2406",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/727",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1660",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 765.23,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 255/878",
   "hardware": "nvidia-gtx-1660",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3ebb66475d2c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/602",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 741.99,
   "pp_test": "LocalScore avg",
   "tg_tps": 31.14,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 258/878",
   "hardware": "nvidia-gtx-1050",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ccc405437425",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/742",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1070 with Max-Q Design",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 813.53,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.39,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 274/878",
   "hardware": "nvidia-gtx-1070-max-q",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "38bff8f90093",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1939",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 582.81,
   "pp_test": "LocalScore avg",
   "tg_tps": 41.02,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 278/878",
   "hardware": "apple-m2-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "599b8bf25aa0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/432",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 583.93,
   "pp_test": "LocalScore avg",
   "tg_tps": 39.81,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 279/878",
   "hardware": "apple-m1-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "4fff2d601b03",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3351",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1650 SUPER",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 666.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 31.17,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 285/878",
   "hardware": "nvidia-gtx-1650-super",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "772d2976fe69",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/17",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 573.86,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.94,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 289/878",
   "hardware": "apple-m1-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "224a83bb3d2c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1281",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+7GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 527.09,
   "pp_test": "LocalScore avg",
   "tg_tps": 38.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 303/878",
   "hardware": "apple-m1-7c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "f42ce9de60b0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/223",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 4P+4E+7GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 530.89,
   "pp_test": "LocalScore avg",
   "tg_tps": 37.87,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 304/878",
   "hardware": "apple-m1-7c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "8c1e437e4bcc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/245",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 4P+4E+8GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 523.21,
   "pp_test": "LocalScore avg",
   "tg_tps": 34.48,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 317/878",
   "hardware": "apple-m2-8c",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "edc6ffdc6793",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1249",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1050",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 538.52,
   "pp_test": "LocalScore avg",
   "tg_tps": 27.4,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 337/878",
   "hardware": "nvidia-gtx-1050",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d8b34095afde",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1582",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GT 1030",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 543.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.05,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 348/878",
   "hardware": "nvidia-gt-1030",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "6f716c953654",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3526",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro P620",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 539.3,
   "pp_test": "LocalScore avg",
   "tg_tps": 24.41,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 349/878",
   "hardware": "nvidia-quadro-p620",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "dc034cba87e0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3128",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple A18 Pro 2P+4E+5GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 429.38,
   "pp_test": "LocalScore avg",
   "tg_tps": 33.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 353/878",
   "hardware": "apple-a18-pro-5-core-gpu",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "b1ff70475536",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/374",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1650 Ti",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 567.17,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.71,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 360/878",
   "hardware": "nvidia-gtx-1650-ti",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "e56fb90ded25",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4259",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Quadro T2000",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 572.15,
   "pp_test": "LocalScore avg",
   "tg_tps": 20.42,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 362/878",
   "hardware": "nvidia-quadro-t2000",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "40f7e5cbb461",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/889",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA T1200 Laptop GPU",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 539.63,
   "pp_test": "LocalScore avg",
   "tg_tps": 21.02,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 364/878",
   "hardware": "nvidia-t1200-laptop",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "dc2048df9269",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/314",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 690.44,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.66,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 367/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c5562bab3202",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1781",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1650",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 526.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.06,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 371/878",
   "hardware": "nvidia-gtx-1650",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "3f1b0c2d2bad",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3251",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1650 Ti with Max-Q Design",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 546.92,
   "pp_test": "LocalScore avg",
   "tg_tps": 19.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 374/878",
   "hardware": "nvidia-gtx-1650-ti-max-q",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "678e6e887351",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/2943",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 631.94,
   "pp_test": "LocalScore avg",
   "tg_tps": 12.05,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 382/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "949b3bf2666c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4062",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 640.33,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 390/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "c1cc1a20733e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1266",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon(TM) 780M",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 617.18,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 395/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "fa5ad11d9070",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1942",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 572.32,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.2,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 418/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "40cf6cee222e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1071",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 591.95,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.28,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 419/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ecc40e69d2b8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3862",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 567.41,
   "pp_test": "LocalScore avg",
   "tg_tps": 11.01,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 427/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "64b99a4a6b36",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3399",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 760M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 480.0,
   "pp_test": "LocalScore avg",
   "tg_tps": 10.37,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 475/878",
   "hardware": "amd-radeon-760m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d346f22e788c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3199",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD RadeonT 780M",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 505.7,
   "pp_test": "LocalScore avg",
   "tg_tps": 7.53,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 496/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "ed48f58fe610",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/234",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce MX150",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 311.46,
   "pp_test": "LocalScore avg",
   "tg_tps": 13.46,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 529/878",
   "hardware": "nvidia-mx150",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d1b1cb0b3daf",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1655",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Radeon RX 5500 XT",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 173.59,
   "pp_test": "LocalScore avg",
   "tg_tps": 56.24,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 564/878",
   "hardware": "amd-radeon-rx-5500-xt",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "d3a7d411c6d3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/4296",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 333.62,
   "pp_test": "LocalScore avg",
   "tg_tps": 9.54,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 575/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "96c6f66345a0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1701",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 399.49,
   "pp_test": "LocalScore avg",
   "tg_tps": 5.91,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 581/878",
   "hardware": "amd-radeon-780m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "48bddd2d44db",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/3391",
   "model_page": "https://www.localscore.ai/model/3",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 740M Graphics",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 211.72,
   "pp_test": "LocalScore avg",
   "tg_tps": 6.21,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 682/878",
   "hardware": "amd-radeon-740m",
   "model": "llama-3-2-1b-instruct",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "9386e18fbde9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "localscore",
   "source_name": "LocalScore (localscore.ai)",
   "source_url": "https://www.localscore.ai/accelerator/1289",
   "model_page": "https://www.localscore.ai/model/4293",
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce GTX 1060 6GB",
   "quant": "Q4_K_M",
   "backend": "llamafile",
   "flash_attn": null,
   "pp_tps": 505.55,
   "pp_test": "LocalScore avg",
   "tg_tps": 22.7,
   "tg_test": "LocalScore avg",
   "commit": null,
   "contributor": null,
   "notes": "LocalScore rank 1/1",
   "hardware": "nvidia-gtx-1060",
   "model": "qwen3-4b-thinking-2507",
   "partial_offload": false,
   "method": "LocalScore",
   "id": "108bdf2828dc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 4505.82,
   "pp_test": "pp2048",
   "tg_tps": 83.43,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gpt-oss-20b-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "78aaba42b446",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2443.91,
   "pp_test": "pp2048",
   "tg_tps": 58.72,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gpt-oss-120b-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "ad24941ed2ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q8_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2986.97,
   "pp_test": "pp2048",
   "tg_tps": 61.06,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Qwen3-Coder-30B-A3B-Instruct-Q8_0-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "qwen3-coder-30b-a3b-instruct",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9153b9f553ca",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q8_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2250.28,
   "pp_test": "pp2048",
   "tg_tps": 29.43,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Qwen2.5-Coder-7B-Q8_0-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "qwen2-5-coder-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "bfc5e70d62ff",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q4_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5948.74,
   "pp_test": "pp2048",
   "tg_tps": 81.05,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gemma-3-4b-it-qat-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gemma-3-4b-it-qat",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "383e4262636c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/dgx-spark/dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q8_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2364.18,
   "pp_test": "pp2048",
   "tg_tps": 48.68,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/GLM-4.7-Flash-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "glm-4-7-flash",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "de5a231f95a8",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 2713.4,
   "pp_test": "pp2048",
   "tg_tps": 129.97,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gpt-oss-20b-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "35cc0804f1fa",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 1648.69,
   "pp_test": "pp2048",
   "tg_tps": 85.6,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gpt-oss-120b-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5c61ba7a7561",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 2453.11,
   "pp_test": "pp2048",
   "tg_tps": 78.97,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Qwen3-Coder-30B-A3B-Instruct-Q8_0-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "qwen3-coder-30b-a3b-instruct",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "0592315d3089",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 1565.91,
   "pp_test": "pp2048",
   "tg_tps": 79.68,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Qwen2.5-Coder-7B-Q8_0-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "qwen2-5-coder-7b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1603fbb33640",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q4_0",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 2923.59,
   "pp_test": "pp2048",
   "tg_tps": 134.28,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/gemma-3-4b-it-qat-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "gemma-3-4b-it-qat",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8cf1f98cb0e9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/mac-m2-ultra/mac-m2-ultra.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra",
   "quant": "Q8_0",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 1629.33,
   "pp_test": "pp2048",
   "tg_tps": 59.58,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/GLM-4.7-Flash-GGUF; chip named only by the file's directory (M2 Ultra), GPU core count not stated",
   "hardware": "apple-m2-ultra",
   "model": "glm-4-7-flash",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "de04b9c4de2d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/nemotron/nemotron-dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q4_K",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 768.84,
   "pp_test": "pp2048",
   "tg_tps": 19.94,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Nemotron-3-Super-120B-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "nemotron-3-super-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "9dfba00df0b3",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-benches",
   "source_name": "llama.cpp benches/ (maintainer runs)",
   "source_url": "https://github.com/ggml-org/llama.cpp/blob/master/benches/nemotron/nemotron-dgx-spark.md",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA DGX Spark (GB10)",
   "quant": "Q8_0",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2761.9,
   "pp_test": "pp2048",
   "tg_tps": 52.85,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "ggml-org maintainers",
   "notes": "Benchmarked repo: https://huggingface.co/ggml-org/Nemotron-3-Nano-4B-GGUF",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "nemotron-3-nano-4b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "26ce3563d0ae",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M3 Ultra 80GPU",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 2816.47,
   "pp_test": "pp2048",
   "tg_tps": 115.52,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m3-ultra-80c",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e6fe60b1ad2f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 76GPU",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 2191.13,
   "pp_test": "pp2048",
   "tg_tps": 116.08,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m2-ultra-76c",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "950557449c52",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M2 Ultra 76GPU",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 1244.57,
   "pp_test": "pp2048",
   "tg_tps": 79.68,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "apple-m2-ultra-76c",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5690c882e8f0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M4 Max",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 1277.42,
   "pp_test": "pp2048",
   "tg_tps": 92.36,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m4-max",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "fa6b9336b4e9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 994.75,
   "pp_test": "pp2048",
   "tg_tps": 75.15,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m1-max",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e565b3e02bc7",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Pro",
   "quant": "MXFP4",
   "backend": "Metal",
   "flash_attn": true,
   "pp_tps": 515.76,
   "pp_test": "pp2048",
   "tg_tps": 45.68,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m1-pro",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "f4ac9a431a0a",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "RTX PRO 6000 Blackwell Max-Q",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9480.55,
   "pp_test": "pp2048",
   "tg_tps": 249.96,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d70663ce7971",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "RTX PRO 6000 Blackwell",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 11521.95,
   "pp_test": "pp2048",
   "tg_tps": 286.91,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "16cfd4f4c4b0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "RTX PRO 6000 Blackwell Max-Q",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 4494.2,
   "pp_test": "pp2048",
   "tg_tps": 170.62,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "960c0f59d791",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "RTX PRO 6000 Blackwell",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5518.07,
   "pp_test": "pp2048",
   "tg_tps": 196.31,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "nvidia-rtx-pro-6000-blackwell",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5de8f0fbeec5",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 3090",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 5170.56,
   "pp_test": "pp2048",
   "tg_tps": 161.77,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-3090",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "d6c38e1c9da9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 8022.33,
   "pp_test": "pp2048",
   "tg_tps": 221.95,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-4090",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3a1673d51602",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4080 SUPER",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 8170.95,
   "pp_test": "pp2048",
   "tg_tps": 186.51,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-4080-super",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "db84ba6a0b89",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5060 Ti",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3839.21,
   "pp_test": "pp2048",
   "tg_tps": 111.51,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5060-ti",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "5f47156c55af",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5070 Ti",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 6339.76,
   "pp_test": "pp2048",
   "tg_tps": 189.45,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5070-ti",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c46a2c1b5a3e",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 7476.55,
   "pp_test": "pp2048",
   "tg_tps": 204.85,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5080",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e8130c6a23b9",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5090",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9848.38,
   "pp_test": "pp2048",
   "tg_tps": 282.51,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5090",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "125a7a1eb93f",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Tesla V100 16GB",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3526.65,
   "pp_test": "pp2048",
   "tg_tps": 117.71,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-tesla-v100-16gb",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "76135309303c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon RX 7900 XT",
   "quant": "BF16",
   "backend": "ROCm,RPC",
   "flash_attn": true,
   "pp_tps": 4251.56,
   "pp_test": "pp2048",
   "tg_tps": 101.92,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "ggerganov",
   "notes": "llama-bench model column: gpt-oss 20B BF16",
   "hardware": "amd-radeon-rx-7900-xt",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "762f5042a502",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14146721",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "Apple M1 Max",
   "quant": "MXFP4",
   "backend": "Metal,BLAS",
   "flash_attn": true,
   "pp_tps": 994.75,
   "pp_test": "pp2048",
   "tg_tps": 75.15,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "0xshivamagarwal",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "apple-m1-max",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2525a946ed57",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14367805",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 6403.85,
   "pp_test": "pp2048",
   "tg_tps": 197.21,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "SoftwareRenderer",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "nvidia-rtx-pro-6000-blackwell-max-q",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "c4a6db153ed2",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15077345",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "AMD Radeon 780M Graphics",
   "quant": "MXFP4",
   "backend": "ROCm",
   "flash_attn": true,
   "pp_tps": 218.99,
   "pp_test": "pp512",
   "tg_tps": 19.35,
   "tg_test": "tg16",
   "commit": null,
   "contributor": "Djip007",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "amd-radeon-780m",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "1e71045a766d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14923651",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1642.09,
   "pp_test": "pp2048",
   "tg_tps": 42.32,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "Djip007",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060",
   "model": "gpt-oss-20b",
   "partial_offload": true,
   "method": "llama-bench",
   "id": "336bc4cce24d",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14923651",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "F16",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1429.64,
   "pp_test": "pp2048",
   "tg_tps": 32.73,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "Djip007",
   "notes": "llama-bench model column: gpt-oss 20B F16; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060",
   "model": "gpt-oss-20b",
   "partial_offload": true,
   "method": "llama-bench",
   "id": "dd6371512540",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14923651",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 333.1,
   "pp_test": "pp2048",
   "tg_tps": 18.85,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "Djip007",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060",
   "model": "gpt-oss-120b",
   "partial_offload": true,
   "method": "llama-bench",
   "id": "f6237f5b8fad",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14923651",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4060",
   "quant": "F16",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 201.31,
   "pp_test": "pp2048",
   "tg_tps": 17.41,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "Djip007",
   "notes": "llama-bench model column: gpt-oss 120B F16; model file is larger than this card's VRAM, so part of it ran on the CPU/system RAM",
   "hardware": "nvidia-rtx-4060",
   "model": "gpt-oss-120b",
   "partial_offload": true,
   "method": "llama-bench",
   "id": "413385f7515b",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15079003",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 3729.54,
   "pp_test": "pp2048",
   "tg_tps": 84.18,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "phoyd",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "06025bc540bc",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15079003",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 1815.6,
   "pp_test": "pp2048",
   "tg_tps": 57.19,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "phoyd",
   "notes": "llama-bench model column: gpt-oss 120B MXFP4 MoE",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "160b63bdd7f0",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15891976",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 12523.67,
   "pp_test": "pp2048",
   "tg_tps": 233.75,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5080",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4cd4872d6207",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15891976",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 11250.7,
   "pp_test": "pp2048",
   "tg_tps": 248.6,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5080",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4cd4872d6207",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15891976",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 9557.96,
   "pp_test": "pp2048",
   "tg_tps": 225.02,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-5080",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "4cd4872d6207",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-15891976",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 5080",
   "quant": "F16",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 12994.53,
   "pp_test": "pp2048",
   "tg_tps": 178.5,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B F16",
   "hardware": "nvidia-rtx-5080",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "2bff61be2115",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-16063223",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 10365.85,
   "pp_test": "pp2048",
   "tg_tps": 254.63,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-4090",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3d241b674db6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-15396",
   "source_name": "llama.cpp discussion #15396",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-16063223",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GeForce RTX 4090",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": null,
   "pp_test": null,
   "tg_tps": 258.02,
   "tg_test": "tg128",
   "commit": null,
   "contributor": "PaulKnowler",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-rtx-4090",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "3d241b674db6",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-16578",
   "source_name": "llama.cpp discussion #16578",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/16578#discussioncomment-14980101",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "MXFP4",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": null,
   "pp_test": null,
   "tg_tps": 86.23,
   "tg_test": "tg128",
   "commit": "c7b7db044",
   "contributor": "eugr",
   "notes": "llama-bench model column: gpt-oss 20B MXFP4 MoE",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "8e4c1b7fc5e1",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-16578",
   "source_name": "llama.cpp discussion #16578",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/16578#discussioncomment-16377832",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 4452.88,
   "pp_test": "pp2048",
   "tg_tps": 86.05,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "am17an",
   "notes": "llama-bench model column: gpt-oss 20B Q4_K - Medium",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-20b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "feea1a30409c",
   "first_seen": "2026-10-21"
  },
  {
   "source_id": "llamacpp-16578",
   "source_name": "llama.cpp discussion #16578",
   "source_url": "https://github.com/ggml-org/llama.cpp/discussions/16578#discussioncomment-16377832",
   "model_page": null,
   "retrieved_at": "2026-10-21T11:38:36Z",
   "hardware_raw": "NVIDIA GB10",
   "quant": "Q4_K_M",
   "backend": "CUDA",
   "flash_attn": true,
   "pp_tps": 2443.59,
   "pp_test": "pp2048",
   "tg_tps": 62.36,
   "tg_test": "tg32",
   "commit": null,
   "contributor": "am17an",
   "notes": "llama-bench model column: gpt-oss 120B Q4_K - Medium",
   "hardware": "nvidia-dgx-spark-gb10",
   "model": "gpt-oss-120b",
   "partial_offload": false,
   "method": "llama-bench",
   "id": "e5fa9be9f7d3",
   "first_seen": "2026-10-21"
  }
 ]
}