Research

Chinese AI Models Reportedly Achieve Perfect Scores on IMO 2026 Problems

Huawei and RedNote say their in-house AI systems solved the 2026 International Mathematical Olympiad problem set with perfect results under the IMO's official judging procedures.

Chinese AI Models Reportedly Achieve Perfect Scores on IMO 2026 Problems

Huawei and RedNote have announced that their in-house artificial intelligence models achieved 100% results on the 2026 International Mathematical Olympiad (IMO) problem set. RedNote — the company behind the Xiaohongshu social platform — stated that its model dots-note-3.0 was the first to complete the contest problems flawlessly according to the IMO's official evaluation procedures.

Both companies emphasized that the AI systems only received the problems after the contestants had finished the exam, and that the solutions had to be submitted within a specified time window. RedNote said no human intervention occurred during the test and that the submissions were evaluated by the Olympiad organizers.

Event details and official results

The 2026 International Mathematical Olympiad took place in Shanghai, with a total of 666 contestants under the age of 20. Official results show that seven participants achieved maximum scores.

Huawei: Celia demonstrated broad problem-solving abilities

Huawei reported that its AI system named Celia showed comprehensive problem-solving capabilities across several mathematical areas. The company did not provide additional internal technical reports in its statement.

Context: recent progress in AI performance

The announcement notes that AI performance has improved significantly in recent years: on the 2025 IMO problems, models from Google and OpenAI already reached results comparable to a gold medal, although they did not achieve perfect scores at that time.

Why this matters

If the companies' claims are accurate, this would indicate that certain large language and problem-solving models can perform at a level of formal mathematical reasoning consistent with the IMO's strict evaluation procedures. The research and education communities are likely to conduct detailed reviews of the models' solutions, the presentation of those solutions, and further specifics of the evaluation process.

The statements in this article are based on announcements from Huawei and RedNote; the companies have not published all detailed review documents from the organizers to the public.