Google has released a major upgrade to its specialized reasoning mode, Gemini 3 Deep Think, designed to address complex challenges in science, research, and engineering. The update is now available for Google AI Ultra subscribers in the Gemini app and via the Gemini API through an early access program.
The upgraded model achieves gold-medal level performance on the International Math Olympiad 2025 and sets new standards on rigorous academic benchmarks. Specific results include a score of 48.4% on Humanity’s Last Exam, 84.6% on ARC-AGI-2 verified by the ARC Prize Foundation, an Elo of 3455 on Codeforces, and gold-medal results on the written sections of the 2025 International Physics and Chemistry Olympiads.
Early testers report that Deep Think successfully identified logical flaws in mathematical papers and optimized fabrication methods for crystal growth. The mode also enables practical applications such as converting sketches into 3D-printable files by modeling complex shapes.