Hardware /
Apple M1 Max (24-core GPU)
18measured results
106best gen tok/s*
32 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 | 106 | 1,684 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 106 | 1,644 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 105 | 1,680 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 105 | 1,627 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 3.2 1B Instruct | Q4_K - Medium | 83.0 | 1,537 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Llama 2 7B | Q4_0 | 54.6 | 400 | llama.cpp (Metal) ↗ | 512+128 | llama.cpp Metal | Details → |
| Llama 2 7B | Q8_0 | 37.8 | 406 | llama.cpp (Metal) ↗ | 512+128 | llama.cpp Metal | Details → |
| Phi 3 Mini 128k Instruct | Q4_K - Medium | 37.1 | 450 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Meta Llama 3.1 8B Instruct | Q4_K - Medium | 32.1 | 306 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Deepseek-R1-0528-Qwen3-8B | Q6_K | 24.5 | 288 | LocalScore ↗ | 512+128 | llamafile 0.9.3 | Details → |
| Llama 2 7B | F16 | 22.6 | 453 | llama.cpp (Metal) ↗ | 512+128 | llama.cpp Metal | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 20.2 | 158 | LocalScore ↗ | 4,352 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 17.6 | 159 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 17.5 | 160 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 17.4 | 156 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 13.8 | 130 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Qwen2.5 14B Instruct | Q4_K - Medium | 11.7 | 112 | LocalScore ↗ | 512+128 | llamafile 0.9.2 | Details → |
| Mistral Small 3.1 24B Instruct 2503 | Q6_K | 7.2 | 74.6 | 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.