How we got this number
Each generated token reads the whole model from memory once: 43 GB. The Mac mini M4 (16GB) moves 120 GB/s, so the ceiling is 2.8 tokens/sec. Real machines reach a fraction of that ceiling; we use 80%, calibrated against a measured Mac run. Send us your real numbers and we'll replace the estimate.
Faster machines for Llama 3.1 70B
| Mac Studio M5 Ultra (96GB) | 23 tok/s | Check price → |
| Mac Studio M5 Ultra (256GB) | 23 tok/s | Check price → |
| Mac Studio M5 Max (40-core GPU, 64GB) | 11 tok/s | Check price → |
| Mac mini M5 Pro (64GB) | 5.7 tok/s | Check price → |
Other models on the Mac mini M4 (16GB)
| DeepSeek Coder v2 16B | Comfortable | 11 tok/s |
| DeepSeek-R1 14B | Comfortable | 11 tok/s |
| Qwen2.5 14B | Comfortable | 11 tok/s |
| Qwen2.5 Coder 14B | Comfortable | 11 tok/s |
| Phi 4 14B | Comfortable | 11 tok/s |
| Qwen 3 14B | Usable, but slow | 6.2 tok/s |