A big moment for AI has arrived—Google’s top model, Gemini with Deep Think enabled, just won a gold medal at the International Math Olympiad (IMO)! Tackling some of the world’s toughest math problems takes serious reasoning and creativity, and Google’s DeepMind team has confirmed the win. IMO President Prof. Dr. Gregor Dolinar shared, “We can confirm that Google DeepMind has hit a major milestone, earning 35 out of 42 points for a gold medal. The solutions were amazing—graders found them clear, precise, and easy to follow.”
This isn’t just about computers being good at math—it’s something new. Gemini uses end-to-end natural language processing, turning problem text into clear mathematical proofs that impressed the graders. The success comes from an upgraded Gemini Deep Think, a smarter reasoning layer that tackles complex questions. It uses a “parallel thinking” approach, exploring multiple solution paths at once and blending them before settling on an answer, moving beyond a straight-line thought process. This shows AI is stepping up to handle self-sufficient, multi-step reasoning—and Gemini did it all within the 4.5-hour limit, meeting the competition’s standards.
We don’t know the exact cost, but running a model like this for 4.5 hours can get pricey, especially with multi-trillion-parameter models on Google TPUs and test time scaling. Soon, Google will roll out the Deep Think model to Gemini Ultra subscribers, who get higher usage limits for $249.99 a month.
What do you think about this AI breakthrough? I’d love to hear your thoughts!
Source: DeepMind
