AWQ
Also known as: AWQ quant
AWQ (Activation-aware Weight Quantization) is a 4-bit scheme that decides which weights to keep at higher precision by looking at activation statistics.
- Keeps the “important” weights precise.
- Fast inference (designed for speed).
- Popular for serving quantized models.
AWQ vs GPTQ: both are strong 4-bit options. AWQ is often preferred for serving speed; GPTQ for broad tooling. Pick one and move on — the gap is small.