Skip to main content
AWQ
conceptQuantization Method
Try in PlaygroundRSS
Overview
Developed byMIT Han Lab
LicenseMIT License
Open source✓ Open Source
Use case4-bit weight quantization for large language models
Also see
Alternative to
Knowledge graph stats
Claims14
Avg confidence92%
Avg freshness100%
Last updatedUpdated 20 days ago
Trust distribution
100% unverified
Governance
EU Risknot classified

AWQ

concept

Activation-aware Weight Quantization method for efficient LLM compression with minimal accuracy loss.

Compare with...

open source

ValueTrustConfidenceFreshnessSources
trueUnverifiedHighFresh1

published year

ValueTrustConfidenceFreshnessSources
2023UnverifiedHighFresh1

stands for

ValueTrustConfidenceFreshnessSources
Activation-aware Weight QuantizationUnverifiedHighFresh1

primary use case

ValueTrustConfidenceFreshnessSources
4-bit weight quantization for large language modelsUnverifiedHighFresh1

reduces

ValueTrustConfidenceFreshnessSources
model memory footprintUnverifiedHighFresh1

developed by

ValueTrustConfidenceFreshnessSources
MIT Han LabUnverifiedHighFresh1

license type

ValueTrustConfidenceFreshnessSources
MIT LicenseUnverifiedHighFresh1

supports model

ValueTrustConfidenceFreshnessSources
LLaMAUnverifiedHighFresh1
OPTUnverifiedHighFresh1

requires

ValueTrustConfidenceFreshnessSources
PyTorchUnverifiedHighFresh1

integrates with

ValueTrustConfidenceFreshnessSources
Hugging Face TransformersUnverifiedModerateFresh1

maintains

ValueTrustConfidenceFreshnessSources
model accuracy during quantizationUnverifiedModerateFresh1

alternative to

ValueTrustConfidenceFreshnessSources
GPTQUnverifiedModerateFresh1
SmoothQuantUnverifiedModerateFresh1

Alternatives & Similar Tools

alternative to
Compare
alternative to
Compare

Commonly Used With

Related entities

Graph Insights

Top sources (14 claims traced)
maintainshighsource
reduceshighsource
published_yearhighsource
requireshighsource
alternative_tohighsource
Trace all provenance
Claim count: 14Last updated: 4/25/2026Edit history