Google’s New Gemini Pro Features Are Now Available, but Most of Them Will Cost You a Pretty Penny.
Google has officially unveiled its Gemini 3 Pro , and for the first time, it’s already available to the market without forcing you to wait months after the announcement to get your hands on it. Along with several new features, Google claims the model boasts improved accuracy as usual, but has also finally gotten rid of the excessive flattery that drives me crazy when using AI. The catch? Most of the features are paid, and one is still in development.
As for more obvious improvements, Google claims the Gemini 3 Pro ranks first in LMArena, scoring 1,501 points, and also demonstrates “PhD-level logical reasoning” on the final Humanity exam, scoring 37.5% without using tools. Math enthusiasts will also be pleased to know that the new model scored 91.9% in GPQA Diamond and 23.4% in MathArena Apex.
But for everyone else, the truly exciting features of Gemini 3 Pro are the all-new features. These aren’t just performance improvements. Gemini 3 Pro isn’t just Gemini 2.5, but better, with three new features for users and a new platform for developers. It’s a much more comprehensive release than, say, ChatGPT 5.1 , if you’re willing to pay more.
Generative user interface
Google’s press release alternates between calling this feature “Generative User Interface” and “Generative Interfaces,” though I prefer the former. Essentially, it’s supposed to make AI output more readable.
It’s one of the few new features that’s available to everyone, although Google has two different approaches to it, and not everyone may like them.
The first option is called a “visual layout” and is more similar to Gemini’s current search results pages. Essentially, when you enter a complex, multi-layered query, such as “plan a three-day trip to Rome next summer,” you’ll receive a visual itinerary rather than static text. It will include photos and interactive modules, but will remain within the familiar Gemini interface. It may also include interactive sliders and buttons for further refinement, but Google says the idea is to create an “immersive, magazine-like experience.”
The second option is more like a web page on demand. It’s called a “dynamic view” and essentially uses agent-based coding to generate an app on the fly that helps you learn more about a topic. It will include generated text and images, but may look very different from the Gemini interface you’re used to. The press release gives an example of how Google generates a response to the dynamic query “explain the Van Gogh Gallery using the life context for each work.” This creates a scrolling page with a clickable title, a left-aligned image, and scrollable text on the right with clickable subheadings and quotes. All this with a custom font and design that can differ significantly from other Gemini results.
One of the challenges I encountered with long AI text responses is that skimming them can be a bit tedious, and Google hopes this will help address that. The company hasn’t yet decided on an approach. Both visual layout and dynamic preview are already available to both free and paid subscribers, but the company states that “to make it easier for us to compare these experiments, you may only see one of them at first.”
Gemini 3 Deep Thinking
Unlike Generative UI, Gemini 3’s Deep Think mode requires a significant paywall and is still under development. This feature is an evolution of Gemini 2.5’s existing Deep Think mode and essentially allows the AI to spend more time answering a question to better consider the appropriate answer. It’s similar to existing free deep thinking modes in Gemini and other AI apps, but has a broader application and, according to Google, is well-suited for tasks like complex graphic design or programming.
Unfortunately, it’s currently only available to a few “security testers.” Google told the press that the rollout will begin “in the coming weeks,” but even then, it will only be available to Google AI Ultra subscribers. Considering Google AI Ultra costs $250 per month (though the starting price is $125 per month for the first three months), this is quite a pricey proposition.
Gemini Agent
Another exclusive Google AI Ultra feature is Gemini Agent, which will begin rolling out to subscribers today. It’s similar to the company’s recent AI-powered search and shopping initiative . The idea is that it will perform actions for you, helping you complete multi-step tasks without leaving Gemini.
The key is that it works with other Google apps. For example, you can ask it to help organize your Gmail inbox, and it will categorize your pending emails and send emails it thinks you can delete for your approval. Or you can ask it to help you rent a car within a certain budget for an upcoming trip, and it will scan your emails for flight and hotel information, then find a suitable reservation and contact you before finalizing it. It has access to the web, Google Workspace, and other AI tools like Canvas , so in theory, it can use virtually all of Google’s resources to help you answer your question.
Google also says it will “ask you to confirm before performing critical actions like making purchases or sending messages, and you can take control at any time.”
If it works, it would be similar to the kind of virtual secretary most people probably envisioned when companies first started talking about AI just a few years ago. But with such a high monthly fee, that’s probably only for the wealthiest for now. So, something like a regular secretary, I guess.
Google Antigravity
Finally, this feature is aimed more at developers than general internet users, but it’s worth mentioning simply because it’s free. Google Antigravity is a new development platform focused on agent-based coding, or code generation powered by artificial intelligence. It’s actually more complicated than that: you can freely view and edit the generated code, but the idea is to make it easier to use Gemini as a development partner.
This isn’t the first Gemini development tool , but the idea is to provide developers with a dedicated AI workspace. To achieve this, Antigravity leverages existing features like Canvas and implements “browser-driven capabilities” and “asynchronous interaction patterns” to achieve an “agent-centric product form factor” that can “autonomously plan and execute complex, end-to-end software tasks.” I’m sure those who use Antigravity will understand what this means. For a non-expert like me, the main improvement is that it’s a single app that lets you go from idea to publication, rather than switching between several different Gemini tools.
But perhaps even more interesting is that Antigravity is free, “with generous restrictions on Gemini 3 Pro.” Yes, developing with Gemini is now actually cheaper than with Gemini, depending on what you want to do.
How to use Gemini 3
At least some of the Gemini 3 Pro’s features are available to everyone. To try the Gemini 3 Pro yourself, open the Gemini app or webpage and select “Thinking” from the model selection menu under your search query. Google announced that free Gemini 3 Pro users will be able to receive up to five queries per day, while AI Plus, Pro, and Ultra subscribers will receive “higher limits.”
Google AI Pro and Ultra subscribers will also be able to use Gemini 3 Pro directly from AI mode in Google Search by selecting “Thinking” in the model selector (next to the “AI mode” button). Unfortunately, free users will have to use the Gemini app for now, but Google says this will change “soon.” On the other hand, AI mode will still be able to generate visual layouts and dynamic views if you have access to Gemini 3 Pro.
AI Overviews will also begin using the Gemini 3 Pro for AI Pro and Ultra subscribers, although this is planned for “the coming weeks.” With the update, AI Mode will automatically send the most difficult questions to the Gemini 3 Pro, without the need to select a model from the selection menu (though you can continue to manually select older models if you prefer).
Personally, I’m glad Google released the Gemini 3 Pro with specific new features, rather than simply making it “AI, but better.” At the same time, since most of the features require a subscription, it’s clear we’re still a long way from the widespread adoption of AI that the tech industry so craves.