Apple M2 Max (38-core GPU) × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on Apple M2 Max (38-core GPU)
25.2best gen tok/s
252best 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 | 25.2 | 252 | LocalScore ↗ | 512+128 | 4,799 ms | llamafile 0.9.2 |
| Q4_K - Medium | 23.8 | 246 | LocalScore ↗ | 512+128 | 4,920 ms | llamafile 0.9.2 |
| Q4_K - Medium | 23.4 | 232 | LocalScore ↗ | 512+128 | 5,299 ms | llamafile 0.9.2 |
| Q4_K - Medium | 20.6 | 226 | LocalScore ↗ | 512+128 | 4,984 ms | llamafile 0.9.2 |
| Q4_K - Medium | 18.8 | 191 | LocalScore ↗ | 512+128 | 6,315 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 M2 Max (38-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.