An artificial intelligence model developed by the Chinese technology company RedNote achieved a milestone by earning a perfect score at this year’s International Mathematical Olympiad (IMO) held in Shanghai. The model, known as dots-note-3.0, successfully solved all six competition problems, securing a full 42 points—a first for AI at this prestigious annual global mathematics contest.
RedNote, based in Shanghai and recognized for its social platform Xiaohongshu, revealed that dots-note-3.0 remains in beta and represents the lightest version in its dots3 AI family. The family also includes larger variants named jazz and aria, which are designed for different use cases and computational requirements. The company indicated plans to release the dots-note-3.0 model as open-source software in the near future, though no specific timeline was provided.
The IMO, which has been held annually since 1959, is considered the most esteemed mathematics competition worldwide, demanding participants to provide not only correct answers but also rigorous, logically sound proofs without errors or omissions. This level of reasoning has posed a significant challenge for artificial intelligence systems, which until now had not attained perfect results.
Previous AI efforts at the IMO have yielded strong but incomplete performances. In the 2023 competition, Google DeepMind and OpenAI each earned gold medal-level scores of 35 out of 42 points. At the 2024 event, DeepMind’s models AlphaProof and AlphaGeometry 2 achieved scores equivalent to the silver medal tier with 28 points. However, these systems required human experts to manually translate the problems into formal logical language before the AI could process and solve them.
In contrast, RedNote’s dots-note-3.0 system processed the original mathematical problems directly from the official competition documents, without any human assistance or rule violations. This capability was made possible by an “agent” framework that integrates natural language understanding with Python code execution, enabling the model to draft, test, and refine its problem proofs iteratively. RedNote attributed the model’s success to its multi-step self-review process, which allows it to critically evaluate its reasoning, challenge underlying assumptions, and correct errors before final submission.
On the human front, the Chinese national team also delivered a strong performance at the 2024 IMO, placing first overall. Among the six team members, three achieved perfect scores.
This development marks a significant advance in the ability of AI systems to tackle complex mathematical reasoning tasks, potentially impacting future approaches to automated problem solving and formal verification.