How we got this number
Each generated token reads the active part of the model from memory once: 5.8 GB. The Mac Studio M5 Ultra (96GB) moves 1228.8 GB/s, so the ceiling is 120+ 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 Qwen3.8 flash next 125B (A6B)
| Mac Studio M5 Ultra (256GB) | 84 tok/s | Check price → |
| NVIDIA DGX Spark (128GB) | 9.8 tok/s | Check price → |
| Ryzen AI Max+ 395 mini PC (128GB) | 9.2 tok/s | Check price → |
Other models on the Mac Studio M5 Ultra (96GB)
| Nemotron 3.5 lightning 30B (A3B) | Fast | 120+ tok/s |
| gpt-oss 120B | Fast | 120+ tok/s |
| mixtral 8X7B | Fast | 82 tok/s |
| Qwen3.5 122B (A10B) | Fast | 48 tok/s |
| Nemotron 3 super 120B (A12B) | Fast | 38 tok/s |
| Nemotron 3 33B | Fast | 35 tok/s |