Hardware /
RTX 4090
38measured results
432best gen tok/s*
24 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 | 432 | 27,767 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 396 | 25,164 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 382 | 25,538 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 282 | 25,807 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 2 7B | Q4_0 | 190 | 10,830 | llama.cpp (Vulkan) ↗ | 512+128 | llama.cpp Vulkan @4ae88d0 — coopmat2 | Details → |
| Llama 2 7B | Q4_0 | 189 | 14,771 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @2241453 | Details → |
| Llama 2 7B | Q4_0 | 188 | 9,452 | llama.cpp (Vulkan) ↗ | 512+128 | llama.cpp Vulkan @4ae88d0 — coopmat2 | Details → |
| Llama 2 7B | Q4_0 | 186 | 11,993 | llama.cpp (CUDA) ↗ | 512+128 | llama.cpp CUDA @2241453 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 164 | 22,081 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 144 | 19,804 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 125 | 7,871 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 122 | 6,657 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 120 | 7,594 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 115 | 6,909 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 103 | 7,398 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 98.5 | 7,159 | LocalScore ↗ | 4,352 | llamafile 0.9.3 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 84.2 | 7,015 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 80.5 | 6,998 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 72.5 | 4,361 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 70.8 | 3,828 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 69.6 | 4,110 | LocalScore ↗ | 4,352 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 68.8 | 4,208 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Model Hf | Q8_0 | 68.1 | 11,064 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 66.1 | 12,782 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 65.6 | 3,809 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 60.3 | 4,106 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 59.7 | 4,122 | LocalScore ↗ | 4,352 | llamafile 0.9.3 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 58.3 | 5,925 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 51.8 | 4,028 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 48.0 | 3,965 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 43.4 | 10,569 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Phi3 | F16 | 41.3 | 6,298 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 31.9 | 3,161 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 31.7 | 2,663 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 25.8 | 4,014 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 22.3 | 6,976 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 16.9 | 2,471 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 13.1 | 1,973 | 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.