AI Models Shine at the 2025 International Math Olympiad: A Look into the Race Between OpenAI and Google DeepMind
The world of artificial intelligence took a significant leap forward as AI models from OpenAI and Google DeepMind both secured gold medals in the 2025 International Math Olympiad (IMO). This renowned competition, highly regarded as one of the most challenging for high school students, reflects the incredible advancements in AI technologies. Both companies announced their remarkable achievements independently, bringing attention to their ongoing rivalry and the implications for the future of AI.
The Significance of the 2025 IMO Achievements
The results from this year’s IMO not only highlight the rapid development of AI capabilities but also mark a pivotal moment in the competition between major AI players. Both OpenAI and Google are racing to establish themselves as leaders in the field, striving for public perception of dominance—a narrative that can significantly influence their appeal to top-tier AI talent. Since many AI researchers have roots in competitive mathematics, victories in prestigious competitions like the IMO carry immense weight within the research community.
Transition from Formal to Informal Evaluation
In contrast to the previous year, when Google received a silver medal with a “formal” system requiring human translation of math problems, both companies showcased their “informal” systems this year. These systems were designed to understand natural language questions directly and generate proof-based answers autonomously. Remarkably, both AI models managed to answer five out of six questions correctly, surpassing the performance of many high school participants and Google’s own 2024 entry.
Researchers’ Insights on AI Reasoning
Researchers from OpenAI and Google shared their viewpoint on the significance of these gold-medal performances in the realm of AI reasoning models, particularly in non-verifiable domains. Historically, AI reasoning has excelled at tasks with clear-cut answers—like basic math problems—but struggles with more subjective questions that may require nuanced understanding, such as selecting the best chair for comfort or conducting intricate research. This achievement suggests that both organizations are making strides toward overcoming these limitations.
Controversy Surrounding Announcements
However, the celebrations were somewhat overshadowed by controversy. Shortly after OpenAI made its announcement, Demis Hassabis, CEO of Google DeepMind, criticized OpenAI for prematurely declaring its results. He argued that OpenAI’s announcement came too soon, before IMO had officially verified the results and awarded accolades to the participating students. This public dispute illustrates the fierce competition and sometimes juvenile tactics employed in the AI race.
More Read
Google’s Ethical Approach
Google’s approach was more measured; the company waited for the proper acknowledgment of the students before unveiling its results. Thang Luong, a senior researcher leading Google DeepMind’s IMO project, emphasized the importance of respecting the participants. Google had been collaborating with the IMO organizers since the previous year and wanted to ensure that its results adhered to established grading guidelines, viewing this as essential for maintaining integrity in the competition.
OpenAI’s Independent Evaluation
On the other hand, Noam Brown, a senior researcher associated with OpenAI’s IMO initiative, mentioned that the company was approached by IMO to participate in a formal evaluation. However, OpenAI decided to focus on its informal capabilities, believing they were more pertinent to their development goals. After achieving gold-medal results, OpenAI asserted that it employed third-party evaluators—comprised of former IMO medalists—who were qualified to assess the performance of its AI model.
Evaluators and Official Grading
Despite the differences in their approaches, both companies sought validation for their accomplishments. While some may argue that Google’s rigorous process lends it more credibility, the underlying narrative remains that AI models from leading labs are evolving quickly. It’s noteworthy that only a small fraction of human competitors performed at par with the AI models, showcasing the significant capabilities of these systems.
The Ongoing Competition in AI
The rivalry between OpenAI and Google DeepMind reflects a broader trend within the AI sector, highlighting that competitive pressure can lead to rapid advancements in technology. OpenAI, once perceived as far ahead in the industry, now faces a closely matched competitor in Google. With expectations of the upcoming release of GPT-5, OpenAI is keen to maintain its standing and public image in the rapidly evolving landscape of artificial intelligence.
Through these unfolding events, the world can witness how the fast-paced advancements in AI continue to shape the future of technology and mathematics, setting the stage for what lies ahead in this captivating arena.
Inspired by: Source
