Google and OpenAI AI Models Win Gold at International Math Olympiad
- tech360.tv
- Jul 22, 2025
- 2 min read
Artificial intelligence models from Google and OpenAI have achieved gold-medal scores at the International Mathematical Olympiad (IMO), marking a milestone in AI's mathematical reasoning capabilities.
This is the first time AI systems have surpassed the gold-medal threshold at the IMO, a prestigious global competition for high-school students. Both companies’ models solved five out of six problems using general-purpose reasoning models that interpret mathematical concepts through natural language.
The 66th IMO was held on the Sunshine Coast in Queensland, Australia, with 630 students participating. Of those, 67 contestants, or about 11%, earned gold medals.
Google’s DeepMind unit used a model called Gemini Deep Think, which was introduced at its developer conference in May. The model completed the problems within the official 4.5-hour time limit using only natural language, avoiding the formal languages and extended computation times used in previous AI approaches.
OpenAI developed an experimental model that scaled up “test-time compute,” allowing the system to think longer and run multiple lines of reasoning in parallel. Researcher Noam Brown described the process as “very expensive” but did not disclose the exact computing cost.
Alexander Wei, an OpenAI researcher, said on social media that the company does not plan to release a model with this level of mathematical capability for several months.
Junehyuk Jung, a mathematics professor at Brown University and visiting researcher at DeepMind, said the achievement suggests AI could soon assist mathematicians in solving unsolved research problems.
The IMO board officially coordinated with AI developers for the first time this year. Judges certified the results and requested that companies publish them on July 28, after student results were verified.
Google DeepMind CEO Demis Hassabis said the company waited to share its results until students had received proper recognition. OpenAI published its results on Saturday after receiving permission from an IMO board member.
Google and OpenAI AI models earned gold medals at the 66th International Mathematical Olympiad
Both models solved five of six problems using natural language reasoning
Google used its Gemini Deep Think model; OpenAI used an experimental high-compute model
Source: REUTERS
