Google has made official what has been expected for weeks: the launch of Gemini 3 Pro, its most ambitious artificial intelligence model to date.
The company has introduced it in an initial availability phase that will give way, in a matter of days, to a global rollout.
What are the new changes introduced by Gemini 3?
However, the presentation was not limited to announcing a new model: it was accompanied by advances affecting the search engine, tools for programmers and, above all, the system’s reasoning capabilities, which Google claims far outstrip those of several of its direct competitors.
Gemini 3 Pro also comes with an enhanced variant, Gemini 3 Deep Think, designed for tasks requiring a deeper level of analysis. The first benchmark results released by Google show scores that place this generation above models such as Claude Sonnet 4.5 or even GPT 5.1, reflecting the effort to lead the race in generative AI.
Generative interfaces: the new way to interact with AI according to Google
One of the announcements that has attracted the most attention is the arrival of generative user interfaces, a feature that aims to change the way people interact with the search engine. Instead of just generating text, Gemini 3 can shape visual and interactive experiences created in real time.
This means that the AI can interpret a query and transform the answer into an interactive map, a physical simulation, an educational quiz or even a travel itinerary with alternative routes. The system analyzes what the user needs, designs a functional interface and generates the necessary code to bring that tool to life in a matter of seconds.
Google claims that this feature will have an immediate impact in areas such as education, web design or daily planning. A student will be able to practice an exam directly from the search engine, an interface designer will be able to validate ideas without generating manual prototypes, and any user will be able to create recipes or step-by-step task lists that the AI will check as they are completed.
Google Antigravity, the most important work of Gemini 3
Gemini 3 Pro also lands with an announcement aimed especially at developers: Google Antigravity, a working environment where AI agents are no longer simple assistants but autonomous collaborators capable of completing complex tasks from start to finish.
While current wizards are often limited to suggesting code or resolving queries, Antigravity opens the door to agents that plan, execute and review tasks on their own, without constant supervision. According to Google, these agents can write functions, run tests, resolve bugs, consult documentation and verify results within the same workflow.
“Now, agents can autonomously plan and execute complex software tasks from start to finish, while validating their own code,” the company explained during the presentation.
This environment, available for Windows, Linux and macOS, integrates technologies such as Gemini 3, Gemini 2.5 Computer Use and Nano Banana, the AI-powered image generator and photo editor. The goal is to move programmers from “doing” to simply defining targets, while agents take care of the rest.
