Gemini 3 is here: 3 things you should know about the major AI update

Google’s new Gemini 3 model aims to improve all of Google’s artificial intelligence functions.
The Gemini app has a new look, smarter reactions, and an agent to tackle more complex online tasks.
The search will use Gemini 3 to generate interactive and visual answers to your most complex questions.

Google has officially unveiled Gemini 3, marking a major update to its artificial intelligence models and the platforms that employ them. New reasoning modes, new app features, and an overhaul of Google’s ecosystem could change the way people interact with Google products in everyday life. The new model arrives with its usual set of benchmarks and leaderboards, but the most interesting part is what it means for those skeptical about whether they should experiment with AI.

Gemini 3 comes in two versions: Gemini 3 Pro and Gemini 3 Deep Think. Pro is the full-featured daily version available immediately in Apps, Search, and Developer Tools. Deep Think is the “enhanced reasoning” mode with additional equipment, currently in testing and aimed at deep-pocketed Google AI Ultra subscribers. But they both share the same core.

1. Gemini Power 3

Google is positioning Gemini 3 as a leap forward in reasoning, not just size or speed. The company likes to talk about AGI’s “steps” in cautious forecasts, but the tone around this statement is clearly more confident. The company wants to boast about how well the Gemini 3 can do everything from analyzing and translating a handwritten recipe to being able to merge it with a voice memo and writing a cookbook from the combination.

That kind of multimodality is where Gemini 3 feels most transformed. The model’s video analysis now understands motion, timing, and other details. It can even analyze a sports match and suggest a training plan for the players. And its million-token context window means you can track extensive real-world information without breaking down mid-way through a long session.

2. Gemini 3 in the Gemini app

Gemini 3 does not arrive in isolation. The Gemini app is getting one of its biggest overhauls, with a new interface navigation system and a “My Stuff” folder for every AI-generated content you’ve ever requested. According to Google, the new app should be more useful by default.

The most notable change is the new generative interfaces. Gemini 3 creates them in real time based on what you request instead of just using a template. As an example, Google suggested that asking for help planning a vacation could produce a magazine-style itinerary complete with images. Or, instead of a wall of text to answer a complicated question, you would see a visually heavy layout of diagrams, tables, and other illustrations.

The Gemini app also has an agent who will act on your behalf. If you give Gemini a task that requires a few dozen steps, it can accomplish it using your connected Google apps. The agent starts with Google AI Ultra members and expands from there.

3. Search for Gemini 3

Gemini 3 is changing the way Google handles complex questions. For the first time, a Gemini model is immediately available in Search, and Google is directing the toughest queries toward it. Google AI Pro and Ultra subscribers in the US will see the Gemini 3 Pro in AI mode, with wider access coming soon

Gemini 3 improves on Google’s distribution approach of searching for multiple interpretations of your question to find relevant content. Gemini 3 understands intent deeply enough to uncover material that previous versions routinely left out.

The most striking update is the new generative user interface. When you ask a complex question, Gemini 3 builds a layout with images, tables, grids, and even custom-coded interactive simulations to accompany the answer. A question about the three-body problem produces a manipulable model. A loan question generates a calculator tailored to the details you’ve included. The answers are more like small applications.

There are also many links to the source material, which in theory encourages following by users. Google says this system will evolve, particularly as automatic model selection sends more queries to Gemini 3 behind the scenes.

Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds. Be sure to click the Follow button!

And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp also.