Google has unveiled a major upgrade to Gemini’s “Deep Think” reasoning mode, designed to tackle complex challenges in science, research and engineering. Developed with input from scientists, the update focuses on solving messy, real-world problems with no single correct answer.
The new Gemini 3 Deep Think shows major benchmark gains, including top scores on Humanity’s Last Exam, ARC-AGI-2, competitive programming (Codeforces), and International Math Olympiad–level tasks, alongside improved performance in fields like physics and chemistry.
Practically, Deep Think can help interpret complex data, model physical systems and even turn sketches into 3D-printable objects. The upgrade is now available in the Gemini app for Google AI Ultra subscribers and via the Gemini API for enterprise users.
Source: 9to5Google
