Apple M3 Pro (18-core GPU) × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on Apple M3 Pro (18-core GPU)
12.3best gen tok/s
151best prompt tok/s
4,352max context tested
9results
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 | 12.3 | 147 | LocalScore ↗ | 4,352 | 8,658 ms | llamafile 0.9.2 |
| Q4_K - Medium | 12.2 | 151 | LocalScore ↗ | 512+128 | 8,546 ms | llamafile 0.9.2 |
| Q4_K - Medium | 12.1 | 149 | LocalScore ↗ | 512+128 | 8,573 ms | llamafile 0.9.2 |
| Q4_K - Medium | 12.1 | 149 | LocalScore ↗ | 512+128 | 8,566 ms | llamafile 0.9.2 |
| Q4_K - Medium | 12.0 | 144 | LocalScore ↗ | 512+128 | 8,953 ms | llamafile 0.9.2 |
| Q4_K - Medium | 11.7 | 147 | LocalScore ↗ | 4,352 | 8,823 ms | llamafile 0.9.2 |
| Q4_K - Medium | 11.7 | 146 | LocalScore ↗ | 512+128 | 8,853 ms | llamafile 0.9.2 |
| Q4_K - Medium | 11.6 | 144 | LocalScore ↗ | 512+128 | 8,970 ms | llamafile 0.9.2 |
| Q4_K - Medium | 11.2 | 141 | LocalScore ↗ | 512+128 | 9,107 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 Pro (18-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.