27 Gems

28490 Points

Badges: Rookie

Google’s Gemini Deep Think Wins Gold at 2025 International Math Olympiad

Google’s advanced AI model, Gemini Deep Think, has achieved a gold-medal-level performance at the 2025 International Mathematical Olympiad (IMO), solving 5 out of 6 problems, a milestone officially confirmed by the competition’s organizers.

Natural Language Understanding: Unlike previous AI systems, Gemini Deep Think solved problems in natural language without requiring human translation into code, working within the 4.5-hour time limit just like human contestants.

Breakthrough Architecture: Its “Deep Think” approach uses parallel reasoning, exploring multiple solution paths simultaneously before selecting the best one, a leap beyond traditional linear AI reasoning.

Creative Problem-Solving: In one case, it found an elegantly simple solution using basic number theory, while human competitors relied on a far more complex graduate-level theorem.

While Gemini stumbled on the hardest problem (solved by only 5 humans), its overall performance marks a major leap in AI’s mathematical reasoning. Google plans to release this model first to trusted testers, including mathematicians, before rolling it out to Google AI Ultra subscribers.