How we got this number
Each generated token reads the whole model from memory once: 19 GB. The Mac mini M5 Pro (24GB) moves 307 GB/s, so the ceiling is 16 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 Code Llama 34B
| Mac Studio M5 Ultra (96GB) | 52 tok/s | Check price → |
| Mac Studio M5 Ultra (256GB) | 52 tok/s | Check price → |
| Mac Studio M5 Max (40-core GPU, 64GB) | 26 tok/s | Check price → |
| Mac mini M5 Pro (64GB) | 13 tok/s | Check price → |
Other models on the Mac mini M5 Pro (24GB)
| lfm 2.5 8B (A1B) | Fast | 120+ tok/s |
| Llama 2 7B | Fast | 65 tok/s |
| Code Llama 7B | Fast | 65 tok/s |
| DeepSeek Coder 6.7B | Fast | 65 tok/s |
| Llama 2 uncensored 7B | Fast | 65 tok/s |
| orca mini 7B | Fast | 60 tok/s |