#159159th of 242 in ML frameworks and libraries

optimum

Hugging Face toolkit for exporting and running models on specific accelerators

Stars
3.5k
License
Apache-2.0
Last commit
Nov 2026
Last release
Aug 2026
Language
Python
Stars, last 29 weeks+181
25 Apr 2026: 3,343 stars2 May 2026: 3,356 stars9 May 2026: 3,362 stars16 May 2026: 3,372 stars23 May 2026: 3,376 stars30 May 2026: 3,382 stars6 Jun 2026: 3,388 stars13 Jun 2026: 3,397 stars20 Jun 2026: 3,405 stars27 Jun 2026: 3,411 stars4 Jul 2026: 3,416 stars11 Jul 2026: 3,425 stars18 Jul 2026: 3,433 stars25 Jul 2026: 3,439 stars1 Aug 2026: 3,444 stars8 Aug 2026: 3,446 stars15 Aug 2026: 3,453 stars22 Aug 2026: 3,458 stars29 Aug 2026: 3,464 stars5 Sept 2026: 3,471 stars12 Sept 2026: 3,478 stars19 Sept 2026: 3,486 stars26 Sept 2026: 3,492 stars3 Oct 2026: 3,495 stars10 Oct 2026: 3,500 stars17 Oct 2026: 3,503 stars24 Oct 2026: 3,508 stars31 Oct 2026: 3,519 stars7 Nov 2026: 3,521 stars14 Nov 2026: 3,524 stars

3,343 on 25 Apr 20263,524 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Optimum extends Transformers, Diffusers, TIMM and Sentence-Transformers with export, quantization and hardware-specific inference and training wrappers. It covers ONNX Runtime, OpenVINO, ExecuTorch, Intel Gaudi, AWS Inferentia/Trainium, NVIDIA TensorRT-LLM, AMD and FuriosaAI through pip extras. Export and optimization run from Python or the optimum-cli.

Who it is for: Engineers deploying Hugging Face models on specific accelerators or edge devices

Strengths

  • Covers many backends: ONNX Runtime, OpenVINO, ExecuTorch, Gaudi, Inferentia, Trainium, TensorRT-LLM
  • Export and optimization available both as Python API and optimum-cli
  • Per-accelerator pip extras, e.g. optimum[openvino], optimum[neuronx]
  • Trainer wrappers for Gaudi and Trainium alongside inference tooling

Weaknesses

  • ONNX support moved to the separate optimum-onnx package, splitting docs and installs
  • Much of the functionality lives in sibling repos such as optimum-intel and optimum-habana
  • Library only; no model server or HTTP API included
  • Each backend needs its own dependencies and hardware, so setup differs per target

What it needs

  • GPU optional
  • Needs Transformers, ONNX Runtime, OpenVINO, ExecuTorch, optimum-onnx
  • Models: Transformers, Diffusers, TIMM, Sentence-Transformers, ONNX

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in ML frameworks and libraries

See all 242
Also in ML frameworks and libraries
RankProjectScore
1transformersModel-definition library for pretrained text, vision, audio and multimodal models169.3k starsApache-2.094 out of 100
2ultralyticsPython library and CLI for training and running YOLO vision models63.5k starsAGPL-3.094 out of 100
3pytorch-image-modelsPyTorch library of image models, pretrained weights and training scripts37.3k starsApache-2.094 out of 100
#4pytorchPython tensor library with GPU support and tape-based autograd105.8k starscustom license93 out of 100
#5supervisionPython toolkit for loading, annotating and counting computer vision detections Live demo ↗ (opens in a new tab)51.7k starsMIT92 out of 100