OpenAI’s experimental large language model (LLM) has made history by achieving gold medal-level performance at the 2025 International Math Olympiad (IMO). Under human exam conditions, the model successfully solved five out of six challenging problems, earning 35 out of 42 points—sufficient to qualify for a gold medal at the prestigious competition. OpenAI researcher Alexander Wei emphasized that the evaluation was conducted under strict guidelines, akin to those followed by human contestants. The IMO is globally recognized for its complexity, requiring extended creative reasoning. The model’s solutions were independently verified by former IMO medallists, showcasing its ability to produce intricate, logical arguments similar to human mathematicians. OpenAI plans to release new advancements soon but clarified that this LLM represents a different, experimental research path. CEO Sam Altman acknowledged this milestone as a significant indicator of AI development, while reflecting on the unexpected levels of progress made since 2021.
Source link
