bitsandbytes Foundation

license

MIT(OSI)

source

public

core features withheld

no

The LICENSE file contains the MIT 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/bitsandbytes-foundation/bitsandbytes/main/LICENSE recorded 2026-09-30

  LICENSE file is the verbatim MIT text, opening with "MIT License" and "Permission is hereby granted, free of charge, to any person obtaining a copy".

• https://raw.githubusercontent.com/bitsandbytes-foundation/bitsandbytes/main/pyproject.toml recorded 2026-09-30

  pyproject.toml declares license = "MIT" and license-files = ["LICENSE"] and builds with "scikit_build_core.setuptools.build_meta" from the repository. Searched the body for enterprise, license key, paid, commercial and proprietary: none.

• https://raw.githubusercontent.com/bitsandbytes-foundation/bitsandbytes/main/README.md recorded 2026-09-30

  README states "`bitsandbytes` is MIT licensed." and lists three features for k-bit quantization with 8-bit optimizers, LLM.int8() and QLoRA 4-bit, with a sponsors section thanking "Hugging Face". Searched the body for enterprise, license key, paid, commercial and proprietary: none.

PyPI downloads of `bitsandbytes` are the adoption signal here.

• https://pypistats.org/api/packages/bitsandbytes/recent recorded 2026-08-18

  last_month downloads = 5,708,239 for bitsandbytes

bitsandbytes provides 8-bit optimizers, LLM.int8() mixed-precision inference and 4-bit quantization for PyTorch across multiple hardware backends. It quantizes existing models rather than compiling or lowering a computation graph, which is what separates it from a full compiler such as Apache TVM.

• https://raw.githubusercontent.com/bitsandbytes-foundation/bitsandbytes/main/README.md recorded 2026-09-30

  README documents "k-bit quantization for PyTorch" with "8-bit optimizers uses block-wise quantization to maintain 32-bit performance at a small fraction of the memory cost." and "QLoRA or 4-bit quantization enables large language model training with several memory-saving techniques".

• https://raw.githubusercontent.com/bitsandbytes-foundation/bitsandbytes/main/README.md recorded 2026-09-30

  README accelerator table covers "NVIDIA GPU", "AMD GPU", "Intel GPU", "Intel Gaudi" and "Metal" plus x86-64 and aarch64 CPU, so multi-backend support is documented; the LLM.int8() entry says "separately treating outliers with 16-bit matrix multiplication".