Google has released a major upgrade to Gemini 3 Deep Think — a specialized reasoning mode built on the Gemini 3 family of AI models — designed specifically to tackle complex scientific, research, and engineering challenges. The update pushes the frontier of machine reasoning, showing significant improvements on rigorous benchmarks and enabling real-world applications like identifying subtle logical flaws in advanced mathematical work, optimizing material fabrication processes, and generating 3D-printable designs from sketches. The new Deep Think is now available in the Gemini app for Google AI Ultra subscribers and, via an early-access API, for researchers, engineers, and enterprises.

Gemini 3 Deep Think: Advancing science, research and engineering