Apple M4 Max (40-core GPU) × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on Apple M4 Max (40-core GPU)
58.8best gen tok/s
704best prompt tok/s
4,352max context tested
14results
Measured · all numbers below come from the linked sources
llama.cpp
| Quant | Gen tok/s | Prompt tok/s | Source | Context tested | TTFT | Build |
|---|---|---|---|---|---|---|
| Q4_K - Medium | 58.8 | 698 | LocalScore ↗ | 4,352 | 1,717 ms | llamafile 0.9.2 |
| Q4_K - Medium | 56.1 | 704 | LocalScore ↗ | 512+128 | 1,721 ms | llamafile 0.9.2 |
| Q4_K - Medium | 55.6 | 672 | LocalScore ↗ | 512+128 | 1,777 ms | llamafile 0.9.2 |
| Q4_K - Medium | 55.5 | 655 | LocalScore ↗ | 512+128 | 1,807 ms | llamafile 0.9.2 |
| Q4_K - Medium | 54.8 | 652 | LocalScore ↗ | 512+128 | 1,812 ms | llamafile 0.9.2 |
| Q4_K - Medium | 54.2 | 663 | LocalScore ↗ | 512+128 | 1,797 ms | llamafile 0.9.2 |
| Q4_K - Medium | 53.1 | 627 | LocalScore ↗ | 512+128 | 1,931 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.4 | 636 | LocalScore ↗ | 512+128 | 1,854 ms | llamafile 0.9.2 |
| Q4_K - Medium | 51.6 | 607 | LocalScore ↗ | 512+128 | 1,994 ms | llamafile 0.9.1 |
| Q4_K - Medium | 51.2 | 590 | LocalScore ↗ | 512+128 | 2,040 ms | llamafile 0.9.2 |
| Q4_K - Medium | 49.6 | 583 | LocalScore ↗ | 4,352 | 2,048 ms | llamafile 0.9.2 |
| Q4_K - Medium | 48.6 | 577 | LocalScore ↗ | 4,352 | 2,093 ms | llamafile 0.9.2 |
| Q4_K - Medium | 43.0 | 526 | LocalScore ↗ | 512+128 | 2,153 ms | llamafile 0.9.2 |
| Q4_K - Medium | 41.8 | 478 | LocalScore ↗ | 512+128 | 2,435 ms | llamafile 0.9.2 |
Ollama
No measured Ollama results yet. Submit yours →
MLX
No measured MLX results yet. Submit yours →
"Context tested" is the largest prompt+generation length actually benchmarked (tokens), not a memory-fit estimate.
What else can Apple M4 Max (40-core GPU) run? →
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.