The Shanghai Artificial Intelligence Laboratory, in collaboration with several leading universities, has released the Olympiad Reasoning Model SU-01, an open-source AI capable of solving International Mathematical Olympiad (IMO) problems without external tools. Trained on P1-30B-A3B, SU-01 scored 35 points on the official IMO 2025 problems, reaching the gold medal cutoff. The model operates using an internal generate-validate-revise loop, demonstrating that complex math problems can be solved without external code executors or theorem provers.
SU-01's success is attributed to its test-time computational expansion, known as Test-Time Scaling (TTS), which allows the model to achieve elite performance through iterative self-refinement. The training involved an inverse perplexity curriculum and reinforcement learning to enhance its reasoning capabilities. Despite its computational intensity, SU-01's ability to self-correct and complete rigorous proofs enables it to perform at a high level, as evidenced by its performance on the 2026 USAMO problems.
Open-Source AI Model Achieves Gold Medal in IMO 2025 Without External Tools
Disclaimer: The content provided on Phemex News is for informational purposes only. We do not guarantee the quality, accuracy, or completeness of the information sourced from third-party articles. The content on this page does not constitute financial or investment advice. We strongly encourage you to conduct you own research and consult with a qualified financial advisor before making any investment decisions.
