Google’s Gemini 3 Deep Think Receives Major Upgrade for Practical Applications
Google has announced a significant enhancement to its Gemini 3 Deep Think, a specialized reasoning mode within the Gemini platform. This upgrade aims to address complex challenges in science, research, and engineering by integrating deep scientific knowledge with practical engineering applications.
Deep Think is designed to tackle intricate research problems that often lack clear solutions and involve incomplete or messy data. By employing advanced reasoning techniques, it explores multiple hypotheses simultaneously, leading to more refined and nuanced outcomes.
The latest upgrade has resulted in notable improvements across various benchmarks:
– Humanity’s Last Exam: Achieved a new standard with a 48.4% score without the use of external tools.
– ARC-AGI-2: Attained an unprecedented 84.6% score, as verified by the ARC Prize Foundation.
– Codeforces: Reached an Elo rating of 3455, showcasing exceptional performance in competitive programming challenges.
– International Math Olympiad 2025: Demonstrated gold-medal level performance, highlighting its advanced mathematical capabilities.
Beyond mathematics and coding, Deep Think has also enhanced its performance in scientific domains such as chemistry and both theoretical and applied physics.
One of the practical applications of this upgrade is in the field of 3D printing. Deep Think can now analyze a sketch, model the complex shape, and generate a file suitable for 3D printing, effectively transforming conceptual designs into tangible objects.
This upgraded version of Gemini 3 Deep Think is currently available to Google AI Ultra subscribers through the Gemini app. Additionally, enterprise users can access this functionality via the Gemini API, with early access available upon request.
