AI from OpenAI and Google DeepMind has reached a new milestone: both companies’ models earned gold-medal scores at the prestigious 2025 International Math Olympiad (IMO), one of the world’s toughest high school-level math competitions. The achievement marks a significant step in the progress of AI reasoning and language models.
From Code to Natural Language: A Shift in AI Testing
This year, both tech giants entered “informal” AI systems that could read IMO questions and generate proof-based answers in natural language without human translation into code or math notation. Each model answered five out of six problems correctly, outperforming most student competitors and even improving on Google’s own 2024 silver-medal performance.
The Race for Perception in AI Talent
These results highlight the rapid development of AI models and the intense competition between OpenAI and Google for industry leadership and user base. Success in highly intellectual benchmarks like IMO matters especially in the AI community, as many researchers come from math competition backgrounds.
Timing and Transparency
However, the celebrations weren’t without controversy. OpenAI announced its gold medal on Saturday, just after the official student results were shared Friday night. Google DeepMind’s leadership pushed back, criticising OpenAI for a premature announcement and for skipping official IMO grading.
Google’s team said they worked closely with IMO organisers, waited for formal grading, and announced their results only after receiving official approval. In contrast, OpenAI had their model’s output reviewed by three former IMO medalists who followed official grading criteria. According to OpenAI, they reached out to IMO after learning of their high score but were advised to wait until after the student awards ceremony to announce.
The Bigger Picture: Rapid AI Progress
While the disagreement over process continues, one point is clear: AI capabilities in reasoning, problem-solving, and natural language understanding are advancing quickly. This year’s IMO saw top AI models match or exceed the scores of elite high school mathematicians, a milestone with real significance in both education and technology.
Final Thoughts
Whether it’s OpenAI or Google, the AI race is tighter than ever. As the field moves toward more complex reasoning tasks, these breakthroughs may soon shift from math contests to solving real-world problems.
- Apple Unveils iPhone 17 Series with New Air Model - September 10, 2025
- Google AI Mode: The Future of Search? - September 10, 2025
- Microsoft Debuts MAI-Voice-1 and MAI-1-Preview - September 9, 2025
