Imagine you're in a global competition to solve the toughest math problems in just 4.5 hours! This is what happened at the International Mathematics Olympiad (IMO) 2025 in Australia, where Google's Gemini Deep Think model surprised everyone by solving 5 out of 6 problems, winning the gold medal with a score of 35/42 points! The organizing committee described his answers as "clear and precise," as if they had been written by a professional mathematician.

Compare this to previous releases like AlphaGeometry and AlphaProof, which took silver last year after solving only 4 issues, so what's the difference? Gemini relied on natural language directly without translation, and completed the solutions in the same time as the competition, while his predecessors took days!

ArabHardware compares the capabilities of artificial intelligence and humans in practical experimentation
The
future of artificial intelligence in Google's grip: everything announced in Google I/O 25

How did Gemini Deep Think become a math genius?

The secret lies in parallel thinking and cutting-edge reinforcement learning techniques, imagine Gemini exploring dozens of solutions at the same time, and then combining them to provide the best answer! The engineers also provided him with a huge database of mathematical problems and their solutions, as well as expert advice on how to think about the Olympics.