Apple M4 Pro (16-core GPU) × Meta Llama 3.1 8B Instruct
Meta Llama 3.1 8B Instruct on Apple M4 Pro (16-core GPU)
30.9best gen tok/s
303best 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 | 30.9 | 302 | LocalScore ↗ | 512+128 | 4,147 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.7 | 291 | LocalScore ↗ | 512+128 | 4,334 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.6 | 302 | LocalScore ↗ | 512+128 | 4,121 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.5 | 303 | LocalScore ↗ | 512+128 | 4,131 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.5 | 302 | LocalScore ↗ | 512+128 | 4,147 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.4 | 292 | LocalScore ↗ | 512+128 | 4,272 ms | llamafile 0.9.2 |
| Q4_K - Medium | 30.2 | 302 | LocalScore ↗ | 512+128 | 4,144 ms | llamafile 0.9.2 |
| Q4_K - Medium | 29.5 | 298 | LocalScore ↗ | 4,352 | 4,218 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 Pro (16-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.