Hardware /
H100 80 GB
6measured results
281best gen tok/s*
79 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 2 7B | Q4_0 | 281 | 11,263 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @5143fa8 | Details → |
| Llama 2 7B | Q4_0 | 268 | 9,918 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @5143fa8 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 96.6 | 14,632 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 95.8 | 14,584 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 42.4 | 4,687 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 23.4 | 2,609 | 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.