Gemini 3 Deep Think: Advancing Science, Research and Engineering

Google announces a major upgrade to Gemini 3 Deep Think, a specialized reasoning mode designed for science, research, and engineering challenges. The updated model sets new benchmark records including 48.4% on Humanity’s Last Exam, 84.6% on ARC-AGI-2, Elo 3455 on Codeforces, and gold-medal performance on the 2025 International Math Olympiad, Physics Olympiad, and Chemistry Olympiad. Early testers demonstrate practical applications: a Rutgers mathematician used it to find a subtle logical flaw missed by human peer review, Duke’s Wang Lab used it to optimize crystal growth fabrication, and a Google R&D lead used it to accelerate physical component design. Deep Think is available to Google AI Ultra subscribers in the Gemini app, with API access opening via an early access program for researchers and enterprises.

Source: Gemini 3 Deep Think: Advancing science, research and engineering