Google has quietly opened access to the Gemini 3 Deep Think API, currently limited to select researchers, engineers, and enterprises through an application form. If you're subscribed to Google AI Ultra, you can start using it now.

Deep Think is a specialized reasoning mode within Gemini designed to tackle 'tough cases'—open-ended problems with incomplete data or ambiguous boundaries. Unlike typical chatbots that provide quick answers, it engages in deep, human-like reasoning to solve problems.

> We've upgraded our specialized reasoning mode Gemini 3 Deep Think to help solve modern science, research, and engineering challenges – pushing the frontier of intelligence. 🧠

>

> Watch how the Wang Lab at Duke University is using it to design new semiconductor materials. 🧵 pic.twitter.com/BgSEmv00JP

> — Google DeepMind (@GoogleDeepMind) February 12, 2026

The results speak for themselves: it scored 48.4% on the benchmark dubbed 'Humanity's Final Exam,' 84.6% on ARC-AGI-2, and boasts an impressive Codeforces Elo rating of 3455. Its strengths lie in complex reasoning for mathematics, physics, chemistry, and engineering.


How does it perform in practice? Duke University's Wang Lab uses it to design novel semiconductor materials; experts have found it capable of spotting overlooked logical errors in research papers; others have fed it sketches and received 3D-printable models in return. Compared to its predecessor, this update covers more disciplines, and API integration allows it to be embedded into existing workflows.

Google is taking a pragmatic approach—eschewing the hype around general AI and instead focusing on solving real-world problems in specific domains. For researchers and engineers, it’s like gaining a partner for 'slow thinking.'

[Source](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think/?ref=testingcatalog.com)