In the weeks leading up to Google I/O 2025, the product name Gemini Ultra popped up a few times, and I thought this would be Google’s next big AI model. We’d either get a Gemini 2.5 Ultra that’s even better than the 2.5 Pro, or the Ultra would be an alternative to the expected Gemini 3 model.
It turns out that Google did plan to unveil a new Gemini AI Ultra product at I/O, but it has nothing to do with the AI technology level we’re at with Gemini. Instead, Gemini AI Ultra is now Google’s most expensive AI subscription, priced at $249.99 per month.
That’s even more expensive than the $200/month ChatGPT Pro subscription OpenAI offers, and for good reason. Gemini AI Ultra includes extra perks in addition to access to Google’s latest AI models and much higher limits. It comes with YouTube Premium and 30TB of storage.
What you get
Google described Gemini AI Ultra as the VIP pass to Google AI in a press briefing BGR attended before the show and in the blogs accompanying the product release. The new subscription tier will give you access to the best of the best AI Google has to offer.
The $250 Gemini AI Ultra subscription includes the following perks, along with the highest usage limits for all of Google’s frontier models:
- Gemini app with full access to models and tools (2.5 Pro Deep Think and Veo 3 will be added soon), plus Gemini Live
- Flow with early access to Veo 3
- Whisk with the highest limits for Whisk Animate
- NotebookLM with the highest limits
- Gemini in Gmail, Docs, Vids & more
- Gemini in Chrome
- Project Mariner
- YouTube Premium
- 30 TB storage
The feature is rolling out on Tuesday in the US, with 70 more countries to follow soon. There’s a 50% discount for the first three months, after which you’ll pay the full $250 monthly rate.
Google also said the existing AI Premium plan will be renamed Google AI Pro. That plan is improving too, with users getting access to Flow video generation (with Veo 2) and early access to Gemini in Chrome.
University students in several countries, including the US, Japan, Brazil, Indonesia, and the UK, will get free access to Google AI Pro for a full year.
The Gemini app improvements
The list of AI features available to Gemini AI Ultra includes many new tools announced at I/O 2025. Google wants the Gemini AI assistant to be “more personal, proactive, and powerful.” That’s why it’s building new features into the Gemini app and giving advanced users better access.
It’s not just the new Gemini Live abilities (which we’ve covered at this link), but also improved models for image and video generation, upgraded Deep Research and Canvas functionality, and support for using Gemini by voice while browsing.
Imagen 4 is Google’s newest AI image generation model. Available in the Gemini app starting Tuesday, it will produce images that “pop with lifelike detail” and offer better text and typography. The images also appear to be watermarked, which is a great feature for genAI content.
Veo 3 is the new AI video generation model coming to Gemini and the Flow AI video-making app. It includes native audio generation, letting you create clips with sound as well as visuals.
Deep Research in Gemini now supports user-provided documentation. This means you can combine your own data with information the AI finds online when creating reports.
Google also upgraded Canvas in Gemini to be more intuitive and powerful. The feature lets you create anything you describe, from infographics and quizzes to Audio Overviews. One of Canvas’s best features is turning ideas into code.
Gemini in Chrome will be available on Wednesday to Windows and macOS users who use US English as their Chrome language and subscribe to Google AI Pro or Gemini AI Ultra. The AI will let you ask Gemini to clarify information or summarize any webpage. Eventually, it will work across tabs and even navigate websites for you.
Finally, Project Mariner is the AI agent Google introduced in December, currently available to a limited group of trusted testers. Mariner can browse the web and make online purchases on your behalf.
Google said at I/O 2025 that Mariner now includes a system of agents that can handle up to 10 tasks at once. They can browse, make bookings, conduct research, and shop online for you. Mariner likely powers shopping features in Gemini Live and Google Search AI Mode.
Project Mariner will initially be available to Gemini AI Ultra users in the US, and its computer-use capabilities will be accessible via the Gemini API.
