AI Systems Match the Best Young Math Minds — and Each Other
Once again, the world of Artificial Intelligence has achieved a significant milestone. Both OpenAI and Google DeepMind announced that their AI models reached gold medal level scoring in the2025 International Math Olympiad (IMO), one of the most prestigious competitions in the world for high school students.
This achievement illuminates the evolution of AI systems. However, the ever-widening gap of the leading AI labs reveals the systems that have been built.
Key takeaways
- OpenAI and Google’s AI models both scored 5 out of 6 IMO problems — a gold medal performance.
- Both employed informal systems that parsed and generated answers in natural language.
- Google chose to wait for official IMO certification before sharing the results.
- OpenAI chose to go earlier and use third-party evaluators.
- Regardless of the arguments, the headline should be how fast AI is getting better at reasoning and solving problems.
What’s the Big Deal About IMO?
The International Math Olympiad is much more than a simple competition for school students. It serves as a measurement of a benchmark high level reasoning – a reasoning that AI systems have always struggled to achieve. Winning the IMO requires much more than arithmetic; one has to demonstrate sophisticated proofs for intricate propositions.
In 2024, Google’s AI systems earned a silver medal, citing a dependency on humans for reformatting questions. A shift occurred this year.
Both OpenAI and Google employed systems with the ability to:
- Parse math problems the same way a student would.
- Create self-contained, thorough, proof-based answers.
- Achieve results that would place them among the best human competitors.
Everyone gets a gold medal, but no clear winner.
Both OpenAI and Google claim to have succeeded and reached the same goals, but the routes they took differ, and that’s where the controversy begins.
What OpenAI Did
- Utilized a reasoning-trained natural language model.
- Brought in three IMO medalists to grade the responses.
- Discovered gold-level performance.
- Sought IMO guidance, subsequently learning there would be a wait.
- Announced results shortly after the student award ceremony.
What Google Did
- Partnered several months earlier with the IMO Organizers.
- Waited for offical grade and validation from IMO’s president.
- Released results only after the student awards were given, per procedure.
This exercise in AI and Perception
Both models performed remarkably, but this goes beyond mathematics. It’s a matter of who is deemed more advanced, more trustworthy, and more respectful of the human competitors.
This is important because:
- In the AI race, the focus is not only on functionalities but also on public image.
- Top math-focused AI researchers usually follow big events like the IMO.
- These “vibes” shape recruitment, funding, and collaboration opportunities.
Currently, the vibe is that Google and OpenAI are in a race against each other in advanced AI development.
What the Results Really Show
Both OpenAI and Google achieved gold medalistic scores. In IMO, gold medal is achieved by a small percentage, so this is quite rare. These AI systems not only passed the test, they also excel far beyond expectations.
Even more impressively, they did it independently. There was no manual effort needed to framework the questions. The questions were in prose, and the systems reasoned, read, and obtained the correct interpretations needed for the solutions.
There has been significant advancement in:
- AI reasoning in elusive or proof-dominated domains.
- Non-verifiable tasks like research or strategic thinking.
- Understanding advanced verbal expressions stripped of formal structures.
What’s Next in the AI Race?
Initially, OpenAI appeared to be in much advanced. It is now evident that Google is catching up rapidly.
Now that GPT 5 is almost releasing, it’s expected OpenAI will regain the lead. On the other hand, Google is now proving that they are able to match performance and follow structured procedure when needed.
Either way, this rivalry is pushing both companies — and AI technology — forward.
Summary: Why It Matters
Final Thoughts
AI is learning faster, and the leaders are closely matched. Even a simple math contest turns into a PR spectacle.
In the words of a certain Team OpenAI member, As the other 99.9% of the population seems to remain unfazed by the awarding of gold medals by OpenAI and Google, Athanassios and I, for the first time in history, seem to wonder why for what would casually qualify a child to the most argued contest in history, Athanassios appears to cope with the rest of the population not knowing the fact that Athanassios and I are laughing, AI has conquered the impossibly challenging test that we the humans so painstakingly devised.