Apple M4 Max (32-core GPU) × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on Apple M4 Max (32-core GPU)
27.3best gen tok/s
309best 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 | 27.3 | 280 | LocalScore ↗ | 512+128 | 4,455 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.9 | 300 | LocalScore ↗ | 512+128 | 4,159 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.9 | 300 | LocalScore ↗ | 4,352 | 4,146 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.8 | 301 | LocalScore ↗ | 4,352 | 4,132 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.7 | 309 | LocalScore ↗ | 512+128 | 4,003 ms | llamafile 0.9.2 |
| Q4_K - Medium | 26.4 | 277 | LocalScore ↗ | 512+128 | 4,465 ms | llamafile 0.9.2 |
| Q4_K - Medium | 22.4 | 235 | LocalScore ↗ | 4,352 | 5,074 ms | llamafile 0.9.2 |
| Q4_K - Medium | 22.3 | 248 | LocalScore ↗ | 512+128 | 4,825 ms | llamafile 0.9.2 |
| Q4_K - Medium | 20.2 | 262 | LocalScore ↗ | 512+128 | 4,303 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.