GPUs & Macs › NVIDIA RTX 3080
NVIDIA RTX 3080: local LLM speed
Llama 2 7B (Q4_0) generates 146 tokens/s on the NVIDIA RTX 3080 (Vulkan), and processes prompts at 4,914 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 146 | 4,914 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 140 | 5,570 | Q4_0 | CUDA FA on | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 140 | 5,014 | Q4_0 | CUDA FA off | tg128 | llama.cpp #15013 details |
| Llama 2 7B | 139 | 4,287 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
| Llama 3.2 1B Instruct | 242 | 11,669 | Q4_K_M | llamafile | LocalScore avg · n=2 | LocalScore LocalScore details |
| Llama 3.1 8B Instruct | 81.1 | 3,025 | Q4_K_M | llamafile | LocalScore avg · n=2 | LocalScore LocalScore details |
| Qwen2.5 14B Instruct | 42.0 | 1,498 | Q4_K_M | llamafile | LocalScore avg · n=2 | LocalScore LocalScore details |
What fits on the NVIDIA RTX 3080
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 146 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | UD-IQ2_XXS | 7.3 GB | 76.7 estimate |
| gemma-4-12B-it | Q5_K_S | 8.2 GB | 67.9 estimate |
| Qwen3.5-9B | Q6_K | 7.5 GB | 74.7 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 124 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q6_K | 7.8 GB | 71.5 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 68.0 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 130 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 82.9 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 464 estimate |
| gemma-4-E2B-it | Q8_0 | 5.0 GB | 112 estimate |
| DeepSeek-R1-0528-Qwen3-8B | Q6_K | 6.7 GB | 82.8 estimate |
| Llama-3.2-3B-Instruct | Q8_0 | 3.4 GB | 163 estimate |
| Phi-4-mini-instruct | Q8_0 | 4.1 GB | 136 estimate |
| gemma-3-4b-it | Q8_0 | 4.1 GB | 135 estimate |
Similar hardware
NVIDIA RTX 2080 Ti · NVIDIA GTX 1080 Ti · NVIDIA RTX 5070 · NVIDIA TITAN V · NVIDIA RTX 2070 SUPER · NVIDIA TITAN Xp · NVIDIA RTX 3060 · NVIDIA RTX 4060 Ti