Chat model by Microsoft · 3.76B parameters
Phi tiny MoE instruct fits comfortably in your 24 GB Mac using Q8_0 quantization, using 20% of your RAM.
~254 tok/s
Q8_0 · 4.7 GB
| Quantization | Memory | Speed | Fits? |
|---|---|---|---|
| Q8_0 Recommended | 4.7 GB | ~254 tok/s | ✓ |
| Q6_K | 3.8 GB | ~333 tok/s | ✓ |
| Q5_K_M | 3.3 GB | ~392 tok/s | ✓ |
| Q4_K_M | 2.9 GB | ~459 tok/s | ✓ |
| Q3_K_M | 2.6 GB | ~555 tok/s | ✓ |
| Q2_K | 2.1 GB | ~720 tok/s | ✓ |
ToolPiper downloads, manages, and runs models with one click. Apple Silicon optimized.
Get ToolPiper — Free