Google’s new AI is now available to all users, but it must be activated manually from the chat’s drop-down menu.

Awesome laptop on sale: 32GB RAM, NVIDIA RTX 5060 and Intel Core 7 chip

Google has just launched Gemini 3 , its new generation of artificial intelligence, available since November 18th. The update comes just eight months after Gemini 2.5 and promises substantial improvements in understanding complex questions. Interestingly, Google has decided to open full access : any user, whether they have a free or paid account, can directly try the Gemini 3 Pro model without restrictions.

Artificial intelligence

The problem is that activating the new version isn’t automatic , as Android Authority explains . By default, the system keeps “Fast” mode selected, which still works with Gemini 2.5. So how do you access the new capabilities? You have to manually switch to the “Thinking” option in the drop-down menu that appears below the question box. This somewhat confusing name masks the fact that “Thinking” is actually Gemini 3 Pro.

Interfaces that are created on the spot and autonomous tools

Among the most striking new features are the generative user interfaces . What does this mean? It means that Gemini 3 can create interactive visual experiences on the fly: personalized maps, physics simulations, or quizzes tailored to the user. This feature is available in the search engine’s AI Mode and represents a significant change from the purely text-based responses of the past.

For developers, Google has unveiled Antigravity , a platform where autonomous agents execute complete programming tasks without constant supervision. These agents have direct access to the editor, terminal, and browser, allowing them to write code, run tests, and debug errors in a continuous workflow . As is typical with major AI announcements , it remains to be seen whether reality will live up to the promises.

The phased rollout means some users still only see the older options (“2.5 Flash” and “2.5 Pro”) instead of the new names. The Deep Research feature still requires an AI Pro subscription, but the core of Gemini 3 Pro is free. In published technical tests, the model achieved 95% in AIME 2025 , surpassing Claude Sonnet 4.5 (87%) and nearly tying with GPT 5.1 (94%) in mathematical reasoning. The results confirm the promised improvements in advanced reasoning capabilities.

This universal openness is a far cry from the launch of Gemini 2.5 Pro, which was initially limited to paying users. Google seems to have learned that mass adoption is more valuable than access restrictions . Now anyone can see if it truly outperforms its competitors or if it remains just empty promises.