Google I/O 2025: All About Google’s New AI Models, Assistants & Subscriptions

Google’s annual developer conference—Google I/O 2025—was held on May 20 and 21, just over a year after the previous year’s edition.

Just as its predecessor, Google I/O 2025 focused majorly on the company’s artificial intelligence (AI) and generative artificial intelligence (GenAI) advancements and offerings.

The event began with two videos showing off Google’s text-to-video capabilities. The first video was a countdown from ten that displayed the numbers appearing in a wide range of situations. The second was set in a western-style town and featured humans, animals, and objects intermingling in the streets and saloons, and ultimately panned towards the sky to reveal the words “Welcome to I/O” in balloons.

Check out the first two minutes.

Once the videos concluded, the stage was taken by Google and Alphabet CEO Sundar Pichai, who delivered a keynote speech, speaking about the company’s current AI progress, new products, and future plans.

Pichai began by saying Google normally doesn’t reveal much in the weeks leading up to I/O, as it would save major announcements for the event. However, in the Gemini era, the company is focused on bringing its best models to users ASAP, which is why they’re “shipping faster than ever.”

This included the company’s rapid pace of development and “relentless model progress,” which comes with some impressive stats.

Google is more than 300 points up in Elo scores, a measure of progress, since the launch of its first-generation Gemini Pro model.

As for Gemini’s progress, it’s used by more than 7 million developers (5x last year’s numbers), the model processes over 480 trillion tokens a month (50x up from last year’s 9.7 trillion tokens), the app has over 400 million monthly active users, and Gemini 2.5 Pro sits atop the LMArena leaderboard across all categories.

Pichai also mentioned the company’s work in the AI infrastructure sector by talking about Google’s seventh-generation TPU—Ironwood—which is the first designed specifically to power thinking and inferential AI workloads at scale.

Coming to Google’s portfolio of AI-powered products, Pichai highlighted various research projects that have enabled powerful products for users.

Google initially debuted Project Starline, a breakthrough 3D video technology, at I/O a few years back. The idea was to make users feel like they were in the same room as someone they were interacting with. At Google I/O 2025, the company introduced Google Beam, a communications platform that combines a video model with an array of six cameras to turn 2D videos into realistic 3D experiences. The first Google Beam devices, built in collaboration with HP, will be available for early customers later this year.

This project also boosted Google Meet, which comes with speech translation capabilities that can match a speaker’s voice, tone, and expressions when translating languages. Google AI Pro and Ultra subscribers can try out (beta) translations in English and Spanish now, with more languages to be added soon. It will also soon roll out to Workspace business customers for early testing.


Google’s Project Astra began as a path to explore the future capabilities of a universal AI assistant capable of understanding the world.

At I/O 2025, Google revealed that Gemini Live uses Project Astra’s camera and screen-sharing capabilities to deliver improved natural interactions across languages, context aware and proactive responses, allow users to enable the AI assistant to take actions on their behalf, summarize what’s important on a screen, retrieve content, complete tasks across numerous apps, and more. All while retaining user preferences to give personalized answers.

Google has also partnered with visual interpreting service Aira to develop and refine Project Astra for blind and low-vision users.

Essentially, the plan is to build an all-inclusive AI assistant that can multitask for users, execute various actions, and offer a diverse range of agentic capabilities.

Speaking about agentic AI capabilities, Google DeepMind’s Project Mariner acts as an early step forward in agents with computer-use capabilities that interact with the web and complete tasks. Google released it as an early research prototype in December and is now bringing its computer use capabilities to developers via the Gemini API.

“Project Mariner now includes a system of agents that can complete up to ten different tasks at a time,” says Google. “These agents can help you look up information, make bookings, buy things, do research and more—all at the same time.”


Building on its move to make Google Search more intelligent, agentic, and personalized, which was done by launching AI Overviews—Google is now introducing the all-new AI Mode, which offers users more advanced reasoning capabilities and allows them to ask longer (even 2x and 3x longer) and more complex queries.

Ahead of this, Google revealed new features such as personalized Smart Replies in Gmail, which can scan through previous emails and Google Drive files to suggest personalized responses.

Google also introduced a comprehensive AI subscription plan called Google AI Ultra. The plan allows users to gain the highest access to our most capable AI models and premium features. It consists of:
 

• The best version of the Gemini app with the highest usage limits across Deep Research, Veo 2’s video generation capabilities, early access to Veo 3, and access to Deep Think in 2.5 Pro.

• Access to Flow, Google’s new custom-designed AI filmmaking tool.

• The highest limits for Whisk Animate, which turns images into eight-second videos.

• Access to the highest usage limits and enhanced model capabilities when using NotebookLM.

• Gemini in Gmail, Docs, Vids, Chrome, and more.

• Project Mariner

• YouTube Premium

• 30TB of storage

Google AI Ultra is now available in the U.S. for $249.99 per month (with a special offer for first-time users of 50% off for the first three months) and is expected to roll out in other countries soon.


What do you think about Google’s AI announcements? Which features/products are you most excited to use?

Let us know in the comments below!