Coding model by codellama · 33.74B parameters
CodeLlama 34b Instruct hf can run but will be tight at 79% of RAM. Consider a lower quantization or closing other apps.
~10 tok/s
Q3_K_M · 18.9 GB
Alternative: Q2_K (15.2 GB, ~14 tok/s)
| Quantization | Memory | Speed | Fits? |
|---|---|---|---|
| Q8_0 | 38.1 GB | ~5 tok/s | ✗ |
| Q6_K | 29.7 GB | ~6 tok/s | ✗ |
| Q5_K_M | 25.7 GB | ~7 tok/s | ✗ |
| Q4_K_M | 22.3 GB | ~9 tok/s | ✗ |
| Q3_K_M Recommended | 18.9 GB | ~10 tok/s | ✓ |
| Q2_K | 15.2 GB | ~14 tok/s | ✓ |
GGUF Sources
ToolPiper downloads, manages, and runs models with one click. Apple Silicon optimized.
Get ToolPiper — Free