Skip to content

Skywork-OR1

Open math and code reasoning models with full training resources

Visit official site

What Skywork-OR1 is

The model series is described as math and code reasoning models developed by the Kunlun Wanwei Tiangong team. It contains three versions for math reasoning, general reasoning and high-performance reasoning tasks, with weights, datasets and training code opened on GitHub and Huggingface.

What you can do with it

  • Solve advanced math problems and contest questions
  • Generate and optimise code automatically
  • Apply the models to education and research fields
  • Work with open model weights, training datasets and training code

Application scenarios

  • A contest participant solves complex math questions with Skywork-OR1-Math-7B
  • A software project generates code snippets with Skywork-OR1-7B-Preview
  • A teacher produces teaching cases with Skywork-OR1-32B-Preview

Getting started

  • Visit GitHub or Huggingface for the open code and training datasets
  • Choose a model version that fits the project needs
  • Install the required libraries and load the model weights locally
  • Run inference tasks such as math problem solving or code generation

Pros & cons

✓ What we like

  • Resources are uploaded to GitHub and Huggingface for reproducible work
  • Training code and datasets are available alongside the weights

! What to watch out for

  • Local use requires installing dependencies and loading model weights
  • A model version has to be chosen to match the task

FAQ

Which models are in the series?

The material lists Skywork-OR1-Math-7B, Skywork-OR1-7B-Preview and Skywork-OR1-32B-Preview.

What is open sourced?

The material states model weights, training datasets and complete training code, uploaded to GitHub and Huggingface.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review