Apple M3 (10-core GPU) × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on Apple M3 (10-core GPU)
1,564best gen tok/s
9,787best prompt tok/s
640max context tested
5results
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 | 1,564 | 9,787 | LocalScore ↗ | 512+128 | 120 ms | llamafile 0.9.2 |
| Q4_K - Medium | 6.3 | 67.3 | LocalScore ↗ | 512+128 | 17,713 ms | llamafile 0.9.2 |
| Q4_K - Medium | 6.1 | 65.2 | LocalScore ↗ | 512+128 | 22,739 ms | llamafile 0.9.2 |
| Q4_K - Medium | 6.0 | 64.9 | LocalScore ↗ | 512+128 | 19,299 ms | llamafile 0.9.2 |
| Q4_K - Medium | 4.3 | 45.8 | LocalScore ↗ | 512+128 | 28,782 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 M3 (10-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.