Google’s new Gemini 3 model brings smarter responses and an AI agent

Today, Google is introducing its newest and most powerful artificial intelligence model, called Gemini 3. The company’s goal with this release is to help people easily “bring any idea to life.” This latest model, starting with Gemini 3 Pro, is now being rolled out for everyone using the Gemini app and the AI Mode feature.

Gemini 3 is designed to be better at understanding what you truly want when you type a request. It is much smarter at reasoning, meaning it can grasp deep and subtle details in complex topics, like scientific ideas or difficult math problems, better than before.

The aim is for Gemini 3 Pro’s answers to be quick, straight to the point, and filled with true insight, making it a great thinking partner for research or creative work. For those who subscribe to AI Ultra, Google will also soon release a stronger version called “Deep Think Mode” for even tougher problems.

One of the biggest changes in this new model is something called “Generative UI.” This means that instead of just giving you text, the AI can actually create and design a fully custom interactive screen or tool right inside the app, built just for your request. For example, if you ask for a trip plan, it might build a travel planning interface, or if you ask about how mortgages work, it could create a working calculator for you to play with.

There are two ways you will see this new feature start to appear. One is called “Dynamic View,” which creates a special, customized, interactive response for whatever you ask. The other, “Visual Layout,” is like a magazine-style page with photos and different sections, and it includes things like sliders and checkboxes that let you quickly change and fine-tune the results right then and there.

A major new feature in the Gemini app, called “Gemini Agent,” is coming to those with an AI Ultra subscription. This experimental tool handles tasks that have multiple steps directly within the app itself. For instance, you could ask the agent to “organize my inbox.”

The agent will then use its smart reasoning and access to tools like Gmail and Google Calendar to group similar emails, let you quickly archive them, or create new reminders in Google Tasks. Another example is asking it to “Book a mid-size SUV for my trip next week using details from my email.”

The agent will find your flight information, search for car rentals that fit your budget, and get the booking details ready. Importantly, the agent will always ask you to confirm before taking critical steps like sending an email or completing a purchase.

Gemini 3 Pro is ready for all users in the Gemini app today; you just need to select the “Thinking” option from the model menu. Users with AI Plus, Pro, and Ultra subscriptions will get higher limits on how often they can use the most powerful settings. This new model is also being added to Google Search and AI Mode, starting with subscribers.

Finally, for software creators, Google announced a new development platform named “Google Antigravity.” This platform is designed to let developers build their own AI agents that can plan and carry out complex software tasks on their behalf across different parts of their computer tools.