Team releases 30B-A3B model achieving gold-medal olympiad performance
Researchers released a 30B-A3B reasoning model that attains gold-medal results on International Physics Olympiad problems and equivalent performance on IMO and USAMO contests. The model applies test-time self-verification for mathematics and a direct approach for physics tasks. It introduces a unified scaling method for automated proof search across both domains. The paper Achieving Gold-Medal-Level Olympiad Reasoning via Simple and Unified Scaling is available on Hugging Face under arXiv 2605.13301.
#68clem 🤗@CLEMENTDELANGUEPaper of the day! https://huggingface.co/papers/2605.13301
Ning Ding@stingningWe’re releasing a 30B-A3B reasoning model that reaches gold-medal level across both physics and math Olympiad evaluations: IPhO directly, and IMO/USAMO with test-time self-verification and refinement. A simple, unified scaling recipe for proof search. https://huggingface.co/papers/2605.13301
#1073Shital Shah@SYTELUSIMO gold at <= 4B active is very real now.
Ning Ding@stingningWe’re releasing a 30B-A3B reasoning model that reaches gold-medal level across both physics and math Olympiad evaluations: IPhO directly, and IMO/USAMO with test-time self-verification and refinement. A simple, unified scaling recipe for proof search. https://huggingface.co/papers/2605.13301