Despite major strides in artificial intelligence, human competitors outperformed AI models developed by Google and OpenAI at the 2025 International Mathematical Olympiad (IMO), held in Queensland, Australia.
For the first time, generative AI systems reached gold-medal-level scores at the prestigious global competition for under-20 mathematicians. However, neither Google’s Gemini nor OpenAI’s experimental reasoning model achieved a perfect score an accomplishment attained by five human contestants who each earned the maximum 42 points.
Google announced that its advanced Gemini chatbot solved five out of the six Olympiad problems, earning 35 out of 42 points enough for a gold medal-equivalent score.
“We can confirm that Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points, a gold medal score,” said 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’s model also scored 35 points, with former IMO medalists independently grading its proof-based answers under competition rules.
Of the 641 student participants from 112 countries, only five human contestants achieved perfect scores, a feat unmatched by any AI system.
This year marks significant progress for AI in mathematical reasoning. In 2024, Google’s system earned only a silver-medal-level score and took two to three days to solve four problems. The latest Gemini model, by contrast, completed its responses within the 4.5-hour time limit allotted to human competitors.
Though tech firms privately tested their models using this year’s actual Olympiad problems, organisers said they could not verify the level of computing power used or whether any human input was involved.
“It is very exciting to see progress in the mathematical capabilities of AI models,” Dolinar added.