In the recently concluded 2025 International Mathematical Olympiad (IMO), AI models from OpenAI and Google performed outstandingly, both winning gold medals. This competition is one of the oldest and most challenging high school mathematics competitions in the world, drawing attention to the intelligence and talent of its participants. Both companies independently announced this achievement, showcasing the rapid development of AI systems and highlighting the intensifying competition between them in the AI field.
Last year, Google won a silver medal in the IMO using a "formal" system, which required humans to convert problems into machine-readable formats. This year, both OpenAI and Google used "informal" systems that could directly process problems and generate natural language proof answers. They claimed that their AI models successfully answered five out of six questions, surpassing most high school students and the AI model used by Google last year.
During interviews, researchers from both companies stated that this gold medal marked a significant breakthrough in AI reasoning models in non-verifiable areas. However, Google raised doubts about OpenAI's score announcement, arguing that OpenAI prematurely declared the result without undergoing official IMO evaluation.
Google DeepMind's CEO and researchers criticized OpenAI on social media, stating that it did not follow the formal procedures of the competition when announcing the results. Google's researchers emphasized that they had communicated fully with the IMO organizers before releasing the scores, respecting the participants.
OpenAI responded that they had hired three former IMO medalists as third-party evaluators to score the model's performance, believing that they fully met the competition's standards.
Although Google's concerns are not without merit, this debate may overshadow a larger context: AI models from top laboratories are rapidly improving. This year, excellent students from around the world participated in the IMO, and the achievements of OpenAI and Google's AI models were only matched by a small fraction of the students, reflecting the rapid advancement of AI technology.
As OpenAI is expected to launch GPT-5 in the coming months, competition in the AI industry will further intensify. Although OpenAI once led, the current situation shows the urgency and intensity of the competition.