Infrastructure / Compilers & Model Optimization

AutoRound

AutoRound is Intel's post-training quantization toolkit for large language and vision-language models, using sign-gradient descent (SignRound) to reach 2-4 bit weights at high accuracy with minimal tuning, across broad CPU, GPU and NPU hardware compatibility.

Verified 2026-09-02 via the GitHub API and the LICENSE body.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
core-gated: ungated

Apache-2.0 license body confirmed. The repository is public and unarchived and builds the whole product, and the README describes no paid tier, enterprise edition or license-gated build beside it, so source is public and the core ungated.

https://github.com/intel/auto-round/blob/main/LICENSE recorded 2026-09-02
LICENSE body carries the verbatim Apache License 2.0 text under an Intel copyright header.
https://api.github.com/repos/intel/auto-round recorded 2026-09-02
Repo metadata for auto-round's canonical repository - private false, archived false, fork false.
https://github.com/intel/auto-round/blob/main/README.md recorded 2026-09-02
README describes AutoRound as an advanced quantization toolkit for LLMs and VLMs achieving high accuracy at 2-4 bit widths through sign-gradient descent, with broad hardware compatibility and no paid tier, enterprise edition or license-gated build beside the published source.

Adoption

3 high confidence

3.0

416,661 PyPI downloads of `auto-round` in the trailing 30 days, which lands in the 100K-1M band of the software usage scale, level 3. The `auto-round-nightly` and `auto-round-hpu` variants the README also offers are not summed, so the figure is a floor.

https://pypistats.org/api/packages/auto-round/recent recorded 2026-09-02
last_month downloads = 416,661 for auto-round

Capability

3 medium confidence

3.0

Banded on the category feature matrix as one optimization family, applied broadly - the same shape as gptqmodel and llm-compressor. Placed two bands below the apache-tvm anchor, on a matrix that bands on how much of the model-to-hardware transformation pipeline a product performs, over how many inputs and targets.

https://github.com/intel/auto-round/blob/main/README.md recorded 2026-09-02
README describes AutoRound as an advanced quantization toolkit for LLMs and VLMs achieving high accuracy at 2-4 bit widths through sign-gradient descent, with broad hardware compatibility and no paid tier, enterprise edition or license-gated build beside the published source.

Verified 2026-09-02