SkyworkAI/Skywork-OR1

SkyworkAI has released the Skywork-OR1 series of open-source reasoning models, including Skywork-OR1, Skywork-OR1-Math-7B, and Skywork-OR1-32B-Preview. These models are designed for mathematical and code reasoning and are trained using large-scale rule-based reinforcement learning. The series includes high-performance models that outperform similarly sized models in both math and coding tasks. Metrics such as AIME24, AIME25, and LiveCodeBench are used for evaluation, with detailed results available in their GitHub repository. Docker and Conda environments for installation are provided, with training scripts coming soon.