# AWQ

Protects the small share of weights that matter most for activations before quantizing the rest to low precision.

## What it is used for

4-bit deployment with strong accuracy retention.
