A Chinese artificial intelligence model, ‘DTS-Note 3.0’, has achieved a perfect score in this year’s International Mathematical Olympiad (IMO), solving all six problems with a total score of 42.
This achievement marks the first time an AI system has achieved a full score and a gold medal-level result in the IMO’s history. DTS-Note 3.0 is now the second AI model to achieve a gold-level result, following Google DeepMind’s ‘Gemini DeepThink’, which scored 35 points in 2025.
How it was achieved
According to Xiaohong Shi, the model used an agent-based reasoning system to generate proofs. The version used in the evaluation is currently limited to internal use but will be made open-source for the AI community soon.
The IMO requires contestants to provide complete proofs for their answers, unlike many other mathematical benchmarks that only require the final answer. Any incorrect step can result in points being deducted, making it a challenging task for both humans and AI models.
The competition
This year’s IMO featured six problems covering algebra, geometry, number theory, and combinatorics, each worth 7 points. According to the organizers, human intervention was strictly prohibited during the evaluation, and the model was not allowed to receive any hints or assistance.
Even for humans, achieving a perfect score is extremely difficult. Only 7 out of over 600 students managed to score 42 points, which is approximately 1% of the participants. The gold medal cutoff this year was 29 points, which the AI model surpassed by 13 points.
Replies
No replies yet. Log in to be the first to reply!