OpenAI and Google DeepMind have both achieved gold-medal level scores at the 2025 International Math Olympiad (IMO), one of the world’s most prestigious and intellectually demanding competitions for high school students. Both companies independently announced that their AI models correctly answered five out of six problems — a feat accomplished by only a small percentage of human participants.
This milestone underscores just how rapidly AI systems are advancing in abstract reasoning and mathematical problem-solving, traditionally considered strongholds of human intelligence.
From Formal Systems to Natural Language Reasoning
Last year, Google participated using a formal system, requiring human translators to convert IMO questions into machine-readable formats. In contrast, the 2025 entries from both OpenAI and Google marked a turning point: they used informal systems that could read and understand the problems in natural language and generate full, proof-based solutions.
This natural language capability reflects a significant step toward general reasoning, enabling AI to tackle problems with less structure and more ambiguity — a crucial advancement for AI applications beyond math.
Dispute Over Disclosure: Timing Matters
While the performance of both models was impressive, controversy arose around how and when results were disclosed.
OpenAI published its gold-level achievement on Saturday morning, shortly after the IMO announced student winners. Google DeepMind quickly criticized the move, claiming that OpenAI had acted prematurely and without official IMO validation.
Thang Luong, a senior researcher at Google DeepMind and lead on the IMO project, told TechCrunch that Google chose to wait until Monday morning to announce its results — after receiving official evaluation and approval from the IMO grading committee.
“The IMO organizers have their grading guideline,” said Luong. “So any evaluation that’s not based on that guideline could not make any claim about gold-medal level [performance].”
Btw as an aside, we didn’t announce on Friday because we respected the IMO Board's original request that all AI labs share their results only after the official results had been verified by independent experts & the students had rightly received the acclamation they deserved
OpenAI defended its approach, stating that it had hired three former IMO medalists to grade its model’s responses using the official scoring rubric. Noam Brown, a senior researcher at OpenAI, added that the company only proceeded with the announcement after being told by IMO organizers to wait until the student ceremony concluded on Friday night.
Interestingly, OpenAI said it was unaware that Google had coordinated directly with IMO to participate in an informal trial this year. The company chose not to participate in last year’s formal test, preferring instead to focus on informal reasoning capabilities.
IMO organizers did not respond to media inquiries regarding the dispute.
More Than a Math Contest — A Symbolic Rivalry
While Google arguably followed the more official and rigorous path, both companies demonstrated that today’s leading AI models are now capable of solving complex mathematical problems at a gold-medal level — a benchmark few humans can reach.
But beyond the numbers, this competition reflects a broader symbolic rivalry. AI companies are battling not just for technical superiority, but for perception, prestige, and talent. Many leading AI researchers come from competitive math backgrounds, so the IMO serves as more than a benchmark — it’s a recruiting signal.
What’s Next in the AI Race?
This outcome suggests that the once-clear lead held by OpenAI may now be narrowing. With the anticipated launch of GPT-5 in the coming months, OpenAI is aiming to reassert its dominance. Meanwhile, Google DeepMind continues to make strategic, research-driven moves in key domains.
Regardless of who takes the lead, one thing is clear: the AI frontier is accelerating, and math — once considered too abstract for machines — is now fair game.
