Gold at the International Mathematical Olympiad
Era 5 · Generative AI, reasoning, and agents (2022–2026) › Alignment and reasoning
- Lineage
- Symbols and logic
- Kind
- Application
Who
Google DeepMind, OpenAI
What happened
An advanced version of Gemini Deep Think and an experimental OpenAI model solve 5 of the 6 problems at the 2025 International Mathematical Olympiad and score 35 of 42 points, gold-medal level. Only Google DeepMind's result is graded and certified by the official coordinators.
Why it mattered
Gold-medal performance at the hardest mathematics competition for students, writing the proofs in natural language and within the exam's time limit.
The full story
Every year the International Mathematical Olympiad brings together the world's best pre-university students, who must prove six problems over two sessions of four and a half hours. Only about 8 percent of participants receive a gold medal. In 2024 Google DeepMind's AlphaProof and AlphaGeometry 2 systems had reached silver-medal level with 28 points, but they needed experts to translate the problems into a formal language and took two to three days.
In 2025 the leap was twofold. On July 19 OpenAI researcher Alexander Wei announced that one of the company's experimental reasoning models, which was not an official entrant, had solved five problems under the same conditions as the students, with no internet or tools, for 35 points. Its solutions were graded by three former Olympiad medalists, not by the official coordinators.
Google DeepMind waited for official certification. On July 21 it announced that an advanced version of Gemini Deep Think had perfectly solved five of the six problems, also for 35 of 42 points, working in natural language from the official problem statements and within the time limit. It was among the first group of systems whose results were graded and certified by the Olympiad's coordinators using the same criteria as for the students. IMO president Gregor Dolinar said the solutions were "clear, precise and most of them easy to follow."
In three years, language models had gone from failing school problems to competing at the elite level of mathematics.
Sources
- Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad (Google DeepMind, July 21, 2025) (opens in a new tab). Thang Luong and Edward Lockhart, 2025
- Google and OpenAI are vying for top AI mathlete (Axios, July 21, 2025) (opens in a new tab). Axios, 2025
- OpenAI Model Wins Gold at International Mathematical Olympiad, or Did It? (TechRepublic, July 2025) (opens in a new tab). TechRepublic, 2025
How to cite
“Gold at the International Mathematical Olympiad.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/international-math-olympiad-gold-2025
This entry was last reviewed on November 13, 2026.