Hardware /
Quadro RTX 6000
10measured results
253best gen tok/s*
22 GBreported memory
*Headline excludes tiny/test models (under 0.6B params or TinyStories-style). Open in rig checker →
llama.cpp
| Model | Quant | Gen tok/s | Prompt tok/s | Source | Context tested | Build | |
|---|---|---|---|---|---|---|---|
| Llama 3.2 1B Instruct | Q4_K - Medium | 253 | 9,138 | LocalScore ↗ | 4,352 | llamafile 0.9.2 | Details → |
| EXAONE-Deep-2.4B | Q4_K - Medium | 105 | 4,158 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 2 7B | Q4_0 | 104 | 3,054 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @b8e09f0 | Details → |
| Llama 2 7B | Q4_0 | 103 | 2,751 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @b8e09f0 | Details → |
| Gemma 3 4b It Abliterated | Q4_0 | 81.3 | 4,079 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 64.9 | 1,891 | LocalScore ↗ | 4,352 | llamafile 0.9.2 | Details → |
| Gemma 3 12b | Q4_0 | 40.1 | 1,384 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Gemma 3 12b | Q4_0 | 39.7 | 1,376 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 37.1 | 1,077 | LocalScore ↗ | 4,352 | llamafile 0.9.2 | Details → |
| Gemma 3 27b | Q4_0 | 22.0 | 647 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
Includes results from LocalScore, a Mozilla Builders project by cjpais, mirrored in the cn0303/local-llm-speed-benchmarks dataset. LocalScore runs use llamafile, which is built on llama.cpp.