LLM Compressor
vLLM (UC Berkeley / vLLM team)LLM Compressor applies quantization and pruning to large language models for deployment with vLLM, covering weight, activation, KV-cache and attention quantization along with algorithms such as SmoothQuant and GPTQ. It reads and writes Hugging Face models, saves in the compressed-tensors format vLLM loads, and supports distributed and disk-offloaded compression of models too large for one device.
Openness
5 high confidence- license
- Apache-2.0(OSI)
- source
- public
- core features withheld
- no
The LICENSE file contains the Apache-2.0 text. The repository is public and the README describes no paid tier, enterprise edition or license-gated build, so the core is ungated.
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/LICENSE recorded 2026-09-30
LICENSE file is the verbatim Apache-2.0 text, opening with "Apache License" and "Version 2.0, January 2004" and "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/README.md recorded 2026-09-30
README presents the library as "the fast, efficient, and easy-to-use library for optimizing models for deployment with vLLM" and installs it with a plain "pip install llmcompressor". Searched the body for enterprise, license key, paid, commercial and proprietar: none.
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/setup.py recorded 2026-09-30
setup.py declares the package license and classifier "License :: OSI Approved :: Apache Software License" and license="Apache 2.0" for the published llmcompressor package.
Adoption
2 high confidenceAdoption is measured from PyPI downloads of the `llmcompressor` package, the name LLM Compressor publishes under on PyPI.
- https://pypistats.org/api/packages/llmcompressor/recent recorded 2026-10-07
27,933 downloads in the trailing 30 days for llmcompressor (pypistats last_month)
Capability
3 medium confidenceLLM Compressor applies weight, activation, KV-cache and attention quantization plus expert pruning to Hugging Face models, emitting compressed-tensors for vLLM. That single optimization family, applied to one deployment target, places it two steps below Apache TVM.
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/docs/index.md recorded 2026-09-30
Docs index says LLM Compressor is "an easy-to-use library for optimizing large language models for deployment with vLLM" and that "Model optimization through quantization and pruning addresses the key challenges of deploying AI at scale".
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/docs/index.md recorded 2026-09-30
Docs index warns "Sparse compression (including 2of4 sparsity) is no longer supported by LLM Compressor due to lack of hardware support and user interest."
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/README.md recorded 2026-09-30
README lists "Comprehensive set of quantization algorithms and transforms for weight, activation, KV cache, and attention quantization" and "DDP and disk offloading support for compressing very large models with hardware efficiency", with "Models saved in the `compressed-tensors` format, compatible with vLLM".
- https://raw.githubusercontent.com/vllm-project/llm-compressor/main/README.md recorded 2026-09-30
README lists "Attention and KV Cache Quantization: FP8, NVFP4" and algorithms "GPTQ", "SmoothQuant" and "REAP expert pruning".
Verified 2026-09-30, except some scores, still awaiting re-confirmation