OpenAI and Google DeepMind Claimed Gold-Level IMO Results in 2025, With Validation Disputes

OpenAI and Google DeepMind said their AI systems achieved gold-medal-level results at the 2025 International Math Olympiad (IMO). Each team scored 35 out of 42 by solving five of the six problems; the sixth was not solved by either model.

The results were validated differently: OpenAI’s performance was checked by former IMO gold medalists, while Google’s was assessed using the organization’s official rubric for that year. The comparability of these methods remains unclear.

The results were described as interesting indicators of AI progress, but their practical significance and the systems’ public value remain debated. Unresolved questions also include whether teams borrowed data or approaches and who made the main contribution to the mathematical results.