Apple M3 Max (30-core GPU) × Qwen2.5 14B Instruct
Qwen2.5 14B Instruct on Apple M3 Max (30-core GPU)
20.9best gen tok/s
242best prompt tok/s
640max context tested
4results
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 | 20.9 | 242 | LocalScore ↗ | 512+128 | 5,092 ms | llamafile 0.9.2 |
| Q4_K - Medium | 20.7 | 236 | LocalScore ↗ | 512+128 | 5,234 ms | llamafile 0.9.2 |
| Q4_K - Medium | 20.5 | 230 | LocalScore ↗ | 512+128 | 5,434 ms | llamafile 0.9.2 |
| Q4_K - Medium | 19.0 | 222 | LocalScore ↗ | 512+128 | 5,380 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 Max (30-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.