Google release major upgrade to Gemini 3 Deep Think

Google has announced a major upgrade to Gemini 3 Deep Think, its most advanced reasoning mode, now optimized to tackle complex challenges in science, research, and engineering. Launched today in a new blog post, the enhancement builds on the model’s earlier successes in mathematics and coding, extending its capabilities into broader scientific domains such as chemistry and physics.

Deep Think employs advanced parallel reasoning, exploring multiple hypotheses simultaneously to deliver deeper, more accurate solutions to intricate problems. This upgrade marks a shift toward real-world applications, positioning the AI as a collaborative tool for researchers and engineers facing data-scarce or highly technical scenarios.

The mode is immediately available to Google AI Ultra subscribers through the Gemini app, where users can select “Deep Think” alongside Gemini 3 Pro. Researchers, engineers, and enterprises can also express interest in early API access via the Gemini API for testing in professional workflows.

Real-world examples highlight the update’s potential. Mathematician Lisa Carbone at Rutgers University used Deep Think to review a technical paper on high-energy physics structures bridging gravity and quantum mechanics.

The model identified a subtle logical flaw that had evaded human peer review. At Duke University’s Wang Lab, researchers leveraged it to optimize crystal growth for semiconductor materials, successfully designing a recipe for thin films larger than 100 μm — a target previous methods struggled to achieve.

Google R&D lead Anupam Pathak tested the system to accelerate physical component design, while a new capability allows users to convert hand-drawn sketches into 3D-printable models by analyzing shapes and generating ready-to-print files.

This evolution underscores Google’s vision of AI as a “force multiplier” for human intellect, handling tedious verification and exploration so experts can focus on creative breakthroughs. As Deep Think continues to advance, it promises to accelerate discovery across disciplines, from fundamental physics to practical engineering innovations.

(source)