OpenAI's experimental language model has achieved gold medal-level performance at the 2025 International Mathematical Olympiad (IMO), solving five out of six problems and earning 35 out of 42 points. This marks a significant advancement in AI's ability to tackle complex mathematical reasoning tasks.
The IMO is a prestigious competition where participants solve challenging mathematical problems under strict conditions. OpenAI's model was evaluated under the same constraints as human contestants, including two 4.5-hour sessions without access to external tools or the internet, and required to provide detailed proofs for each problem.
Alexander Wei, a research scientist at OpenAI, highlighted the achievement, stating that the model can now "craft intricate, watertight arguments at the level of human mathematicians." This accomplishment underscores the rapid progress in AI's reasoning capabilities, moving beyond narrow, task-specific methodologies to general-purpose reinforcement learning and test-time compute scaling.
While this model demonstrates impressive capabilities, OpenAI does not plan to release it or similar models with this level of mathematical proficiency for several months. The upcoming GPT-5 is expected to be released soon but will not feature the same advanced mathematical problem-solving abilities as the IMO-winning model.
This development follows previous AI performances in mathematical competitions, such as Google DeepMind's AlphaProof and AlphaGeometry 2, which solved four out of six problems from the IMO in 2024, achieving a score equivalent to a silver medalist. OpenAI's recent success at the IMO signifies a substantial leap in AI's ability to perform complex mathematical reasoning tasks, setting new benchmarks for machine intelligence in this domain.