Google has officially launched the Gemini 3 Deep Think mode in the Gemini app, available to Google AI Ultra subscribers. This new mode significantly enhances reasoning capabilities, designed to tackle complex mathematical, scientific, and logical problems that even the most advanced models face challenges with.

image.png

Gemini 3 Deep Think has performed exceptionally well in multiple rigorous benchmark tests, such as achieving a score of 41.0% on the "Human Last Exam" without using tools, and reaching an unprecedented 45.1% on the ARC-AGI-2 test when using code execution. This outstanding performance is attributed to its advanced parallel reasoning technology, which can explore multiple hypotheses simultaneously. This mode is an improvement over Gemini 2.5 Deep Think, which recently achieved gold-standard results in the International Mathematical Olympiad and the International University Programming Contest World Finals.

Ultra subscribers can immediately experience the powerful features of Gemini 3 Deep Think by selecting the "Deep Think" mode in the prompt bar and choosing Gemini 3 Pro in the model dropdown menu.

Blog: https://blog.google/products/gemini/gemini-3-deep-think/

Key Points:

- 🧠 The Gemini 3 Deep Think mode is officially launched, enhancing reasoning capabilities and focusing on complex problems.

- 📊 Outstanding performance in strict benchmark tests, scoring 41.0% without using tools and reaching 45.1% when using code.

- 🚀 Ultra subscribers can experience this powerful mode by simply selecting it, driving advancements in AI technology.