How we got this number
Each generated token reads the whole model from memory once: 2.2 GB. The GEEKOM A6 (Ryzen 7 6800H, 32GB) moves 76.8 GB/s, so the ceiling is 35 tokens/sec. Real machines reach a fraction of that ceiling; we use 60%, an assumption until we get measured runs for this kind of machine. Send us your real numbers and we'll replace the estimate.
Faster machines for Phi 3 3.8B
Other models on the GEEKOM A6 (Ryzen 7 6800H, 32GB)
| Qwen3.5 0.8B | Fast | 46 tok/s |
| DeepSeek-R1 1.5B | Fast | 42 tok/s |
| Qwen 1.8B | Fast | 42 tok/s |
| Llama 3.2 1B | Fast | 35 tok/s |
| Qwen 3 1.7B | Fast | 33 tok/s |
| Granite 3.1 moe 1B | Fast | 33 tok/s |