AI hits perfect 42/42 on International Mathematical Olympiad problems
For years, achieving a perfect score at the world's most prestigious mathematics competition was reserved for a handful of exceptionally gifted students. This year, however, artificial intelligence (AI) has also achieved a perfect score for the first time, AFP reported.
Two leading Chinese technology companies, Huawei and Xiaohongshu, announced that their AI models each earned a 100% score on the problems from the International Mathematical Olympiad (IMO), held in July, CE Report quotes AGERPRES.
These are the first AI laboratories to publicly disclose their results, though other companies around the world may follow in the coming days.
"We are thrilled with this result because achieving a perfect score at the IMO is extremely difficult," Xiaohongshu said this week. The company entered the competition with its dots-note-3.0 model.
"To date, no large language model"—the technology behind AI tools such as chatbots—"has ever achieved a perfect score under the IMO's official evaluation process," the company added. Xiaohongshu is best known for operating the RedNote social media platform.
In 2025, AI models developed by Google and OpenAI reached gold medal-level performance but did not match the achievement of the five human contestants who earned perfect scores that year.
This year, 666 human contestants from around the world competed at the IMO, which was held in Shanghai, China. According to the official results, seven participants achieved perfect scores.
The Olympiad, hosted in a different city each year, is open to participants who are under 20 years old.
Huawei also announced on Wednesday that its AI system Celia demonstrated "comprehensive problem-solving capabilities" across multiple areas of mathematics.
Technology companies received the competition problems only after the human contestants had completed the exam, and they were required to submit their answers within a specified deadline.
"During the testing, any form of human intervention was strictly prohibited," Xiaohongshu said, adding that the AI-generated solutions were submitted to the organizers for official grading. This was the company's first participation.
AI's progress has been rapid. At the 2024 IMO, Google achieved a silver medal-level result by solving four of the six problems over the course of two to three days.
Deedy Das, a partner at the U.S. investment firm Menlo Ventures, said he also tested this year's IMO problems on four state-of-the-art AI models.
The models—developed by OpenAI, Anthropic, startup Axiom Math, and Chinese company Moonshot AI with its Kimi K3 model—all scored 42 out of 42 points, Das wrote on LinkedIn.
"The AI frontier has officially surpassed IMO mathematics," he said.
Photo: Wikipedia
