GPUs & Macs › AMD Radeon RX 7900 XTX

AMD Radeon RX 7900 XTX: local LLM speed

Llama 2 7B (Q4_0) generates 191 tokens/s on the AMD Radeon RX 7900 XTX (Vulkan), and processes prompts at 3,890 tokens/s.

Measured results

ModelGenerate
tok/s		Prompt
tok/s		QuantBackendTestSource
Llama 2 7B1913,890Q4_0Vulkan FA ontg128llama.cpp #10879 details
Llama 2 7B1833,727Q4_0Vulkan FA offtg128llama.cpp #10879 details
Llama 2 7B1703,874Q4_0ROCm FA ontg128llama.cpp #15021 details
Llama 2 7B1673,552Q4_0ROCm FA offtg128llama.cpp #15021 details
Llama 3.2 1B Instruct78.44,089Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details
Llama 3.1 8B Instruct42.3875Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details
Qwen2.5 14B Instruct26.8490Q4_K_Mllamafile LocalScore avg · n=2LocalScore LocalScore details

What fits on the AMD Radeon RX 7900 XTX

Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 191 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.

ModelBest quant that fitsFileExpected tok/s
Qwen3.8-27BUD-Q6_K22.0 GB33.2 estimate
gemma-4-12B-itQ8_012.7 GB57.6 estimate
Qwen3.5-9BQ8_09.5 GB76.6 estimate
Qwen3.5-4BQ8_04.5 GB163 estimate
Qwen3.6-35B-A3BUD-Q4_K_S20.9 GB204 estimate
gemma-4-26B-A4B-itUD-Q5_K_M21.2 GB112 estimate
MiMo-V2.6-Distill-Qwen-9BQ8_09.5 GB76.5 estimate
gemma-4-E4B-itQ8_08.2 GB89.1 estimate
Muse-Glimmer-30BUD-Q5_K_M19.2 GB38.0 estimate
Qwen3-4BQ8_04.3 GB170 estimate
gemma-4-31B-itQ5_K_M21.7 GB33.7 estimate
gemma-4-26B-A4B-it-qat-q4_0-unquantizedUD-Q4_K_XL14.2 GB167 estimate
gemma-4-12B-it-qat-q4_0-unquantizedUD-Q4_K_XL6.7 GB109 estimate
gemma-3-1b-itQ8_01.1 GB608 estimate

Open in the picker

Similar hardware

AMD Radeon RX 7900 XT · AMD Radeon RX 9070 XT · AMD Radeon RX 9070 · AMD Radeon RX 7800 XT · AMD Radeon RX 7900 GRE · AMD Radeon RX 6900 XT · AMD Radeon RX 6800 · AMD Radeon VII