Gemini 3 Deep Think Achieves Professional Research-Level Proficiency

Google AI has released a major enhancement to Gemini 3 Deep Think, its advanced reasoning platform designed to tackle complex scientific and engineering problems. Built in collaboration with researchers, the updated version excels at addressing open-ended challenges with incomplete or messy data, blending deep theoretical knowledge with practical problem-solving. The upgrade is available today in the Gemini app for Google AI Ultra subscribers and, for the first time, through the Gemini API for selected researchers, engineers, and enterprises.

Gemini 3 Deep Think continues to set new benchmarks in reasoning and academic performance. It achieves top-level results in mathematics and competitive programming, including gold-medal-level performance at the 2025 International Math Olympiad, elite scores on Codeforces, as well as outstanding results on Humanity’s Last Exam and ARC-AGI-2. The platform now also demonstrates excellence in scientific fields, earning gold medal-level scores on the written sections of the 2025 International Physics and Chemistry Olympiads, while showing strong capabilities in advanced theoretical physics.

Beyond its academic achievements, the update is designed for practical applications. Researchers can analyze complex datasets, and engineers can model physical systems through code. Early access via the Gemini API enables laboratories to explore next-generation materials. This is already being done, for example, by the Pratt School of Engineering at Duke University. In this way, Google AI highlights the practical impact of Deep Think alongside its academic accomplishments.