DeepMind and OpenAI Achieve Gold at ‘Coding Olympics’ in AI Milestone
Both companies’ AI models earned gold-medal scores at the International Mathematical Olympiad, solving five out of six problems in natural language within the time limit
In July 2025, artificial intelligence models from Google DeepMind and OpenAI achieved a gold-medal level performance at the International Mathematical Olympiad (IMO), the world’s most prestigious math competition for high-school students.
Both systems solved five out of six problems, reaching a gold-medal threshold under the same conditions as human contestants.
DeepMind’s model, Gemini Deep Think, officially had its solutions graded by IMO judges and scored thirty-five out of forty-two points.
OpenAI’s experimental model matched that score, though its evaluation was done independently.
Unlike earlier AI systems that relied on formal languages or heavy computation outside strict contest constraints, both models processed mathematical concepts using natural language and completed their tasks within the official four-and-a-half-hour exam period.
DeepMind emphasised that its model’s solutions were not only correct but also clear, precise, and easy to follow for many of the problems.
The achievement represents a significant advance in machine reasoning, demonstrating creative problem-solving in mathematics that approaches the level of top human contestants.
Experts observed that this breakthrough could open new opportunities in scientific and engineering fields where mathematics and logical reasoning are central, from designing better computer chips to advancing drug discovery.
While concerns remain about computational cost and whether contest performance fully translates to real-world applications, the milestone shows that AI’s reasoning capabilities are improving markedly and may soon support cutting-edge research work in multiple disciplines.
Both systems solved five out of six problems, reaching a gold-medal threshold under the same conditions as human contestants.
DeepMind’s model, Gemini Deep Think, officially had its solutions graded by IMO judges and scored thirty-five out of forty-two points.
OpenAI’s experimental model matched that score, though its evaluation was done independently.
Unlike earlier AI systems that relied on formal languages or heavy computation outside strict contest constraints, both models processed mathematical concepts using natural language and completed their tasks within the official four-and-a-half-hour exam period.
DeepMind emphasised that its model’s solutions were not only correct but also clear, precise, and easy to follow for many of the problems.
The achievement represents a significant advance in machine reasoning, demonstrating creative problem-solving in mathematics that approaches the level of top human contestants.
Experts observed that this breakthrough could open new opportunities in scientific and engineering fields where mathematics and logical reasoning are central, from designing better computer chips to advancing drug discovery.
While concerns remain about computational cost and whether contest performance fully translates to real-world applications, the milestone shows that AI’s reasoning capabilities are improving markedly and may soon support cutting-edge research work in multiple disciplines.
