Apple M2 (8-core GPU) × Llama 3.2 1B Instruct
Llama 3.2 1B Instruct on Apple M2 (8-core GPU)
52.9best gen tok/s
717best prompt tok/s
640max context tested
7results
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 | 52.9 | 710 | LocalScore ↗ | 512+128 | 1,817 ms | llamafile 0.9.2 |
| Q4_K - Medium | 52.8 | 717 | LocalScore ↗ | 512+128 | 1,814 ms | llamafile 0.9.2 |
| Q5_K - Medium | 43.5 | 662 | LocalScore ↗ | 512+128 | 1,966 ms | llamafile 0.9.3 |
| Q4_K - Medium | 39.1 | 632 | LocalScore ↗ | 512+128 | 1,908 ms | llamafile 0.9.2 |
| Q4_K - Medium | 24.6 | 429 | LocalScore ↗ | 512+128 | 3,272 ms | llamafile 0.9.2 |
| Q4_K - Medium | 21.3 | 322 | LocalScore ↗ | 512+128 | 3,893 ms | llamafile 0.9.2 |
| Q4_K - Medium | 17.8 | 330 | LocalScore ↗ | 512+128 | 4,280 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 (8-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.