Apple M4 Max (32-core GPU) × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on Apple M4 Max (32-core GPU)
48.7best gen tok/s
584best prompt tok/s
4,352max context tested
8results
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 | 48.7 | 584 | LocalScore ↗ | 512+128 | 2,114 ms | llamafile 0.9.2 |
| Q4_K - Medium | 48.7 | 584 | LocalScore ↗ | 512+128 | 2,113 ms | llamafile 0.9.2 |
| Q4_K - Medium | 48.3 | 580 | LocalScore ↗ | 512+128 | 2,125 ms | llamafile 0.9.2 |
| Q4_K - Medium | 47.7 | 562 | LocalScore ↗ | 4,352 | 2,182 ms | llamafile 0.9.2 |
| Q4_K - Medium | 47.5 | 553 | LocalScore ↗ | 512+128 | 2,220 ms | llamafile 0.9.2 |
| Q4_K - Medium | 47.2 | 551 | LocalScore ↗ | 512+128 | 2,201 ms | llamafile 0.9.2 |
| Q4_K - Medium | 42.2 | 475 | LocalScore ↗ | 4,352 | 2,466 ms | llamafile 0.9.2 |
| Q4_K - Medium | 33.7 | 432 | LocalScore ↗ | 4,352 | 2,593 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 (32-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.