# casper-hansen/AutoAWQ

AutoAWQ implements the AWQ algorithm for 4-bit quantization with a 2x speedup during inference. Documentation:

Repository: https://github.com/casper-hansen/AutoAWQ
Canonical: https://ross.abutalabs.com/products/autoawq
Homepage: https://casper-hansen.github.io/AutoAWQ/
Language: Python
License: MIT
License Family: permissive
Archived: true
Last push: 2025-05-11T08:52:59+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 21, release rhythm 40, longevity 78
- inputs: {"age_days": 1104, "days_push": 479, "days_rel": 479, "gap_med": 18, "n_releases_24m": 6}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2349, forks 307 (observed 2026-08-28T04:06:40.186023+00:00)

## What it is
AutoAWQ is a Python library implementing the Activation-aware Weight Quantization (AWQ) algorithm for 4-bit quantization of large language models, offering faster inference and reduced memory usage versus FP16. It is now officially deprecated and has been adopted by the vLLM project's llm-compressor.

## Use cases
- quantize llm to 4-bit
- run quantized language models on gpu
- reduce vram usage for llm inference
- convert huggingface models to awq format
- export quantized models to gguf
- run llm inference on amd rocm or cpu

## When to choose
- you need to quantize or run existing AWQ 4-bit models with older Torch/Transformers versions
- you want fast 4-bit LLM inference on NVIDIA, AMD, or Intel CPU hardware

## When to avoid
- you are starting a new project - use vLLM's llm-compressor or MLX-LM instead
- you need compatibility with the latest Transformers versions
- you expect bug fixes or ongoing maintenance

## Facets
- artifact type: library
- maturity: abandoned
- function: llm-inference, machine-learning, gpu-computing
- domain: large-language-models, machine-learning, deep-learning
- platform: python, windows
- tags: quantization, 4-bit, awq, huggingface, transformers, deprecated, model-compression, gpu, linux, macos

## Member repositories
- casper-hansen/AutoAWQ (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:40.186023+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:36:57.436746+00:00, confidence not recorded.
  - readme: https://github.com/casper-hansen/AutoAWQ (fetched 2026-08-28T04:06:40.186023+00:00, sha aa8563d8bea3)
  - homepage: https://casper-hansen.github.io/AutoAWQ/ (fetched 2026-08-29T10:17:28.905233+00:00, sha bf25bcb1e0c9)
  - registry_pypi: https://pypi.org/pypi/autoawq/json (fetched 2026-08-29T10:17:28.914594+00:00, sha af70e9721f73)
- Data as of 2026-08-30T08:39:29.467469+00:00.