Apple M4 Pro (20-core GPU) × Llama 3.2 1B Instruct
Llama 3.2 1B Instruct on Apple M4 Pro (20-core GPU)
123best gen tok/s
2,242best prompt tok/s
4,352max context tested
22results
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 | 123 | 2,242 | LocalScore ↗ | 512+128 | 555 ms | llamafile 0.9.2 |
| Q4_K - Medium | 123 | 2,151 | LocalScore ↗ | 512+128 | 567 ms | llamafile 0.9.2 |
| Q4_K - Medium | 122 | 2,134 | LocalScore ↗ | 512+128 | 573 ms | llamafile 0.9.2 |
| Q4_K - Medium | 121 | 2,123 | LocalScore ↗ | 512+128 | 578 ms | llamafile 0.9.2 |
| Q4_K - Medium | 121 | 2,137 | LocalScore ↗ | 4,352 | 572 ms | llamafile 0.9.3 |
| Q4_K - Medium | 120 | 2,150 | LocalScore ↗ | 512+128 | 575 ms | llamafile 0.9.2 |
| Q4_K - Medium | 120 | 2,115 | LocalScore ↗ | 512+128 | 564 ms | llamafile 0.9.2 |
| Q4_K - Medium | 119 | 2,126 | LocalScore ↗ | 512+128 | 569 ms | llamafile 0.9.2 |
| Q4_K - Medium | 119 | 2,211 | LocalScore ↗ | 512+128 | 564 ms | llamafile 0.9.2 |
| Q4_K - Medium | 119 | 2,197 | LocalScore ↗ | 512+128 | 567 ms | llamafile 0.9.2 |
| Q4_K - Medium | 119 | 2,062 | LocalScore ↗ | 512+128 | 575 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,068 | LocalScore ↗ | 512+128 | 579 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,164 | LocalScore ↗ | 512+128 | 568 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,116 | LocalScore ↗ | 4,352 | 575 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,135 | LocalScore ↗ | 512+128 | 567 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,128 | LocalScore ↗ | 512+128 | 580 ms | llamafile 0.9.2 |
| Q4_K - Medium | 118 | 2,119 | LocalScore ↗ | 512+128 | 568 ms | llamafile 0.9.2 |
| Q4_K - Medium | 117 | 2,163 | LocalScore ↗ | 512+128 | 570 ms | llamafile 0.9.2 |
| Q4_K - Medium | 117 | 2,108 | LocalScore ↗ | 512+128 | 576 ms | llamafile 0.9.2 |
| Q4_K - Medium | 117 | 2,133 | LocalScore ↗ | 512+128 | 575 ms | llamafile 0.9.2 |
| Q4_K - Medium | 116 | 2,071 | LocalScore ↗ | 512+128 | 590 ms | llamafile 0.9.2 |
| Q4_K - Medium | 115 | 2,109 | LocalScore ↗ | 512+128 | 581 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 Pro (20-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.