MUMBAI: AI has long been crunching numbers, now it is chasing perfect scores. Chinese technology companies Huawei and Xiaohongshu have claimed that their artificial intelligence models achieved full marks in the 2026 International Mathematical Olympiad (IMO), a feat that, if independently verified, would mark a major milestone in AI’s ability to tackle complex mathematical reasoning.
According to an AFP report, both companies said their AI systems solved all six problems in this year’s Olympiad and received perfect scores under the competition’s official evaluation process. If confirmed, it would be the first time AI has matched the highest score achieved by the world’s best human contestants at the prestigious mathematics competition.
The 2026 International Mathematical Olympiad, held in Shanghai, brought together 666 students aged under 20 from around the world. Only seven participants secured perfect scores, highlighting the exceptional difficulty of the examination.
Huawei said its Celia AI model demonstrated advanced reasoning across multiple branches of mathematics. Meanwhile, Xiaohongshu said its dots-note-3.0 model completed all six problems without human assistance.
The company added that its AI received the questions only after the human competition had concluded and submitted its answers within the official time limit for an independent assessment.
The claimed breakthrough follows rapid advances in AI reasoning over the past year. In 2025, leading models from global AI developers achieved scores comparable with gold medal-level performances at the Olympiad but fell short of obtaining full marks.
This year’s claims suggest frontier AI systems are closing the gap with the world’s strongest young mathematicians in solving highly complex, structured problems.
The race is no longer confined to China. AI developers in both China and the United States have increasingly turned to academic benchmarks such as mathematics, coding and scientific reasoning to demonstrate improvements in their models as competition intensifies.
Industry observers have also reported similar performances from other leading systems. Independent researchers and investors have tested models developed by OpenAI, Anthropic, Moonshot AI and specialist start-up Axiom Math, with several reportedly achieving perfect scores on Olympiad-level problems.
While mathematical competitions represent only one measure of AI capability, the latest results point to significant progress in structured reasoning, an area widely regarded as critical for applications ranging from scientific research and software development to increasingly autonomous AI systems.
Even so, the claims from Huawei and Xiaohongshu are yet to undergo broader independent validation. If verified, they would represent another significant milestone in the rapidly accelerating race to build AI systems capable of matching, and increasingly rivaling, human expertise in some of the world’s most demanding intellectual challenges.
