# AutoGPTQ/AutoGPTQ

An easy-to-use LLMs quantization package with user-friendly apis, based on GPTQ algorithm.

Repository: https://github.com/AutoGPTQ/AutoGPTQ
Canonical: https://ross.abutalabs.com/products/autogptq
Language: Python
License: MIT
License Family: permissive
Topics: transformers, deep-learning, inference, large-language-models, llms, nlp, pytorch, quantization, transformer
Archived: true
Last push: 2025-04-11T13:27:20+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 16, release rhythm 8, longevity 88
- inputs: {"age_days": 1239, "days_push": 509, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5070, forks 543 (observed 2026-08-28T04:09:09.324689+00:00)

## What it is
AutoGPTQ is a Python library for quantizing large language models using the GPTQ weight-only quantization algorithm, with user-friendly APIs integrated into Hugging Face Transformers, Optimum, and PEFT. The project is now unmaintained, with users directed to its successor GPTQModel.

## Use cases
- quantize an LLM to 4-bit with GPTQ
- run GPTQ-quantized models on consumer GPUs
- reduce VRAM usage for llama inference
- load int4 quantized transformers models
- benchmark quantized vs fp16 inference speed

## When to choose
- you need to quantize or run GPTQ models on Linux or Windows with CUDA
- you want Hugging Face Transformers integration for GPTQ models

## When to avoid
- starting a new project - use the maintained GPTQModel fork instead
- you need macOS support or support for the latest model architectures
- you need bug fixes or ongoing maintenance

## Facets
- artifact type: library
- maturity: abandoned
- function: llm-inference, machine-learning, deep-learning
- domain: large-language-models, deep-learning, machine-learning
- platform: python, windows
- tags: quantization, gptq, transformers, pytorch, model-compression, unmaintained, natural-language-processing, linux, gpu

## Member repositories
- AutoGPTQ/AutoGPTQ (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:09.324689+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 2026-08-29T18:02:48.614059+00:00, confidence not recorded.
  - readme: https://github.com/AutoGPTQ/AutoGPTQ (fetched 2026-08-28T04:09:09.324689+00:00, sha 9e7d5c0b904b)
- Data as of 2026-08-30T08:39:29.467469+00:00.