TLDR: Human mathematicians, particularly five young individuals, achieved perfect scores at the 2025 International Mathematical Olympiad (IMO), surpassing advanced AI models from Google and OpenAI. While AI models reached gold-level scores for the first time, they could not match the flawless performance of the top human contestants, highlighting the continued edge of human cognitive abilities in complex problem-solving.
In a significant development at the 2025 International Mathematical Olympiad (IMO) held in Queensland, Australia, human mathematicians demonstrated superior problem-solving capabilities, outperforming sophisticated artificial intelligence models developed by tech giants Google and OpenAI. While both Google’s Gemini chatbot and OpenAI’s experimental reasoning model achieved impressive gold-level scores, they fell short of the perfect scores attained by five young human participants.
The 66th edition of the IMO, a prestigious annual competition for mathematicians under 20, featured 641 young talents from 112 countries. Among them, five individuals achieved the maximum possible score of 42 points. In contrast, Google DeepMind’s advanced Gemini chatbot scored 35 out of 42 points, a score that still qualifies for a gold medal. OpenAI’s experimental reasoning model also secured a gold-level score of 35 points.
This marks a notable milestone for AI in mathematics, as it is the first time AI systems have crossed the gold-medal scoring threshold at the IMO. Google stated that its Gemini chatbot successfully solved five out of the six problems presented at the Olympiad. IMO president Gregor Dolinar, as cited by Google, remarked on the AI solutions, stating, “Their solutions were astonishing in many respects. IMO graders found them to be clear, precise and most of them easy to follow.”
OpenAI researcher Alexander Wei confirmed that their models were evaluated under the same rigorous rules as human contestants, with three former IMO medalists independently grading the AI-submitted proofs. While AI’s performance signifies a breakthrough, particularly in using general-purpose “reasoning” models that process mathematical concepts using natural language, the perfect scores achieved by human contestants underscore areas where human intuition, creativity, and complex problem-solving still hold an advantage.
Also Read:
- AI Breakthrough: Google and OpenAI Models Achieve Gold Medal Performance at International Math Olympiad
- Tech Giants Intensify AI Talent War: Meta and Microsoft Secure Top DeepMind Researchers
Junehyuk Jung, a math professor at Brown University and visiting researcher in Google’s DeepMind AI unit, suggested that this achievement brings AI less than a year away from being a collaborative tool for mathematicians to tackle unsolved research problems. Jung, an IMO gold medalist himself in 2003, believes that “the moment we can solve hard reasoning problems in natural language will enable the potential for collaboration between AI and mathematicians,” a concept that could extend to other scientific fields like physics.
