For the first time, artificial intelligence has attained a flawless result at the International Mathematical Olympiad (IMO), a challenge previously reserved for exceptionally gifted students. Tech firms Huawei and Xiaohongshu reported that their respective AI models solved every problem correctly at this year’s event in Shanghai.
Xiaohongshu noted that their model, dubbed dots-note-3.0, achieved this milestone under strict conditions where human intervention was banned. Huawei’s system, Celia, was also credited with showing advanced problem-solving skills. While large language models had reached gold-medal standards in 2025, they had yet to equal the peak performance of human participants until now.
The progress in this field has been rapid, moving from a silver-medal level in 2024 to a perfect score today. Investor Deedy Das confirmed that models from OpenAI, Anthropic, Axiom Math, and Moonshot AI also managed to earn a maximum score of 42 points on the same set of questions, signaling that the current generation of AI has surpassed the requirements of the competition.
