
The International Mathematical Olympiad: A Premier Challenge
The International Mathematical Olympiad (IMO) stands out as one of the world’s most esteemed global competitions for young mathematicians. Annually, six pre-university students from each participating nation come together to tackle a set of six highly complex mathematical problems. A mere 8% are awarded the illustrious gold medal for their exceptional solutions.
AI Competes at the IMO: OpenAI vs. Google
This year’s IMO brought a modern twist as leading tech giants OpenAI and Google entered the fray with their state-of-the-art language learning models (LLMs).Over the weekend, OpenAI announced a remarkable achievement, claiming its experimental reasoning LLM performed at a level typically reserved for gold medalists. To verify this, OpenAI tested its model on the 2025 IMO problems, adhering strictly to the same conditions as the human competitors. The model was required to solve complex problems during two 4.5-hour exam sessions without the aid of external tools or internet access and was tasked with crafting natural language proofs.
We achieved gold medal-level performance 🥇 on the 2025 International Mathematical Olympiad with a general-purpose reasoning LLM! Our model solved world-class math problems—at the level of top human contestants. A major milestone for AI and mathematics.https://t.co/u2RlFFavyT
— OpenAI (@OpenAI) July 19, 2025
Google’s Gemini Deep Think: A Strong Contender
In a parallel development, Google recently announced that its advanced Gemini Deep Think model successfully solved five out of the six IMO problems, also achieving performance on par with gold medal standards. This enhanced iteration of Gemini integrates cutting-edge reasoning modes drawn from Google’s latest research breakthroughs, including techniques that enable concurrent exploration of multiple problem-solving paths, rather than the traditional sequential approach.
According to the Google DeepMind team, the advancements in Gemini Deep Think are notable:
“To make the most of the reasoning capabilities of Deep Think, we additionally trained this version of Gemini on novel reinforcement learning techniques that can leverage more multi-step reasoning, problem-solving and theorem-proving data. We also provided Gemini with access to a curated corpus of high-quality solutions to mathematics problems, and added some general hints and tips on how to approach IMO problems to its instructions.”
Future Availability of Advanced Models
Google plans to roll out this new version of the Deep Think model to a select group of trusted testers shortly. Following this phase, it will be accessible to Google AI Ultra subscribers, marking a significant step forward in AI’s application in solving complex mathematical challenges.
Leave a Reply