The International Mathematical Olympiad (IMO) is the world’s most prestigious competition for young mathematicians, held annually since 1959. Each participating country sends six elite pre-university mathematicians to solve six exceptionally difficult problems in algebra, combinatorics, geometry, and number theory. Medals are awarded to the top half of contestants, with about 8% receiving a gold medal.
Recently, the IMO has become an aspirational challenge for AI systems to test advanced mathematical problem-solving and reasoning. Last year, Google DeepMind’s AlphaProof and AlphaGeometry 2 systems achieved the silver-medal standard, solving four out of six problems and scoring 28 points. This breakthrough showed AI was approaching elite human mathematical reasoning.
This year, an advanced version of Gemini with Deep Think solved five out of six IMO problems perfectly, earning 35 total points and achieving gold-medal level performance. IMO President Prof. Dr. Gregor Dolinar confirmed the milestone, noting that the solutions were clear, precise, and easy to follow. This marks a significant advance over last year’s result, as the model operated end-to-end in natural language, producing rigorous proofs directly from problem descriptions within the 4.5-hour competition limit.
The achievement was powered by Gemini Deep Think, an enhanced reasoning mode that incorporates parallel thinking, allowing the model to explore multiple solutions simultaneously. Training included novel reinforcement learning techniques and access to curated mathematical solutions. A version of this model will be available to trusted testers before rolling out to Google AI Ultra subscribers.
Google DeepMind continues to collaborate with the mathematical community, seeing this as just the start of AI’s potential. By combining natural language fluency with rigorous reasoning, such agents could become invaluable tools for advancing human knowledge toward AGI.