The AI IMO score has reached a historic milestone after artificial intelligence systems achieved perfect marks at the International Mathematical Olympiad (IMO), matching the highest-performing human contestants for the first time.
Chinese technology companies Huawei and Xiaohongshu announced that their AI models each scored 100 per cent on the problems used in this year’s competition, marking a significant advancement in AI reasoning and mathematical problem-solving. The achievement means artificial intelligence has now reached a level previously reserved for only a handful of the world’s brightest young mathematicians.
Xiaohongshu said its model, dots-note-3.0, secured a perfect score under the IMO’s official evaluation process. “We are delighted with this result, because achieving a perfect score at the IMO is extremely challenging,” the company said in a statement. The company noted that no large language model had previously recorded a perfect result under the competition’s official judging standards.
Huawei also announced that its AI system, Celia, demonstrated strong problem-solving abilities across several branches of mathematics, allowing it to achieve full marks. The AI IMO score comes just one year after major AI systems first reached silver-medal performance and months after several models attained gold-level scores without matching the competition’s top human performers.
The 2026 International Mathematical Olympiad, held in Shanghai, attracted 666 contestants from countries around the world. According to the official results, only seven human participants achieved a perfect score.
The IMO is regarded as the world’s most prestigious mathematics competition and is open only to contestants under the age of 20. This year’s results place AI systems alongside the highest-performing students in one of the most demanding academic competitions globally.
According to Xiaohongshu, technology companies received the competition questions only after human contestants had completed the examination. The firms were also required to submit solutions within a specified time limit.
The company said no human assistance was allowed during testing and that all answers were submitted directly to organisers for official grading. This requirement was intended to ensure the integrity of the evaluation process and provide a fair comparison between human and artificial intelligence performance.
AI performance in advanced mathematics has improved dramatically in a short period. At the 2024 IMO, Google’s AI system reportedly achieved a silver-medal level result by solving four of the six problems over several days.
By 2025, AI models developed by companies including Google and OpenAI had reached gold-medal scores, although they still fell short of the perfect performances recorded by a small number of human competitors. The latest AI IMO score suggests that frontier AI systems have now surpassed one of the most difficult benchmarks in competitive mathematics.
Deedy Das, a partner at Menlo Ventures, stated that he independently tested this year’s IMO questions using four advanced AI systems. According to him, models from OpenAI, Anthropic, Axiom Math, and Moonshot AI’s Kimi K3 each achieved the maximum 42 out of 42 points.
Reacting to the results, Das said the latest performances indicate that the capabilities of advanced AI systems have moved far beyond the traditional IMO benchmark. The perfect AI IMO score is expected to intensify discussions about the future role of artificial intelligence in science, research, education and complex problem-solving.
