At this year’s International Mathematical Olympiad in Queensland, Australia, human participants outperformed advanced AI models developed by Google and OpenAI—though both artificial intelligence systems achieved gold-medal level scores for the first time in the competition’s history.
While five young contestants earned perfect scores of 42 points, neither Google’s Gemini chatbot nor OpenAI’s experimental reasoning model managed to solve all six problems. Google confirmed its AI solved five problems, scoring 35 out of 42 points—a gold-medal threshold matched by roughly 10% of human participants.
“We can confirm that Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points—a gold medal score,” the company stated, quoting IMO president Gregor Dolinar. “Their solutions were astonishing in many respects. IMO graders found them to be clear, precise and most of them easy to follow.”
OpenAI also announced its model achieved a gold-level 35 points. Researcher Alexander Wei called the result a breakthrough, writing on social media: “This achieved a longstanding grand challenge in AI at the world’s most prestigious math competition.”
He emphasized rigorous testing conditions: “We evaluated our models on the 2025 IMO problems under the same rules as human contestants. For each problem, three former IMO medalists independently graded the model’s submitted proof.”
The performance marked significant progress from 2024, when Google’s AI earned a silver medal at the Bath competition by solving four problems—a task requiring days of computation. This year, Gemini generated solutions within the 4.5-hour time limit imposed on human participants.
A total of 641 students from 112 countries competed alongside privately tested AI systems. Dolinar noted the excitement around AI’s advancing capabilities but cautioned: “Contest organizers could not verify how much computing power had been used by the AI models or whether there had been human involvement.”
