Google I/O 2025, Google’s greatest developer convention of the yr, takes place Tuesday and Wednesday on the Shoreline Amphitheatre in Mountain View. We’re on the bottom bringing you the newest updates from the occasion.
I/O showcases product bulletins from throughout Google’s portfolio. We’ve bought loads of information referring to Android, Chrome, Google Search, YouTube, and — in fact — Google’s AI-powered chatbot, Gemini.
Google hosted a separate occasion devoted to Android updates: The Android Present. The corporate introduced new methods to seek out misplaced Android telephones and different objects, further device-level options for its Superior Safety program, safety instruments to guard towards scams and theft, and a brand new design language referred to as Materials 3 Expressive.
Listed below are all of the issues introduced at Google I/O 2025.
Gemini Extremely
Gemini Extremely (solely within the U.S. for now) delivers the “highest level of access” to Google’s AI-powered apps and companies, in line with Google. It’s priced at $249.99 per thirty days and consists of Google’s Veo 3 video generator, the corporate’s new Circulation video modifying app, and a strong AI functionality referred to as Gemini 2.5 Professional Deep Assume mode, which hasn’t launched but.
AI Extremely comes with increased limits in Google’s NotebookLM platform and Whisk, the corporate’s picture remixing app. AI Extremely subscribers additionally get entry to Google’s Gemini chatbot in Chrome; some “agentic” instruments powered by the corporate’s Mission Mariner tech; YouTube Premium; and 30TB of storage throughout Google Drive, Google Pictures, and Gmail.
Deep Assume in Gemini 2.5 Professional
Deep Assume is an “enhanced” reasoning mode for Google’s flagship Gemini 2.5 Professional mannequin. It permits the mannequin to think about a number of solutions to questions earlier than responding, boosting its efficiency on sure benchmarks.
Google didn’t go into element about Deep Assume works, but it surely may very well be much like OpenAI’s o1-pro and upcoming o3-pro fashions, which seemingly use an engine to seek for and synthesize the most effective resolution to a given drawback.
Deep Assume is on the market to “trusted testers” through the Gemini API. Google stated that it’s taking further time to conduct security evaluations earlier than rolling out Deep Assume broadly.
Veo 3 video-generating AI mannequin
Google claims that Veo 3 can generate sound results, background noises, and even dialogue to accompany the movies it creates. Veo 3 additionally improves upon its predecessor, Veo 2, when it comes to the standard of footage it may well generate, Google says.
Veo 3 is on the market starting Tuesday in Google’s Gemini chatbot app for subscribers to Google’s $249.99-per-month AI Extremely plan, the place it may be prompted with textual content or a picture.
Imagen 4 AI picture generator
In line with Google, Imagen 4 is quick — quicker than Imagen 3. And it’ll quickly get quicker. Within the close to future, Google plans to launch a variant of Imagen 4 that’s as much as 10x faster than Imagen 3.
Imagen 4 is able to rendering “fine details” like materials, water droplets, and animal fur, in line with Google. It might deal with each photorealistic and summary types, creating photos in a variety of side ratios and as much as 2K decision.
Each Veo 3 and Imagen 4 will likely be used to energy Circulation, the corporate’s AI-powered video instrument geared in the direction of filmmaking.
Gemini app updates
Google introduced that Gemini apps have greater than 400 month-to-month lively customers.
Gemini Reside’s digital camera and screen-sharing capabilities will roll out this week to all customers on iOS and Android. The characteristic, powered by Mission Astra, lets individuals have near-real time verbal conversations with Gemini, whereas additionally streaming video from their smartphone’s digital camera or display to the AI mannequin.
Google says Gemini Reside will even begin to combine extra deeply with its different apps within the coming weeks: It would quickly have the ability to provide instructions from Google Maps, create occasions in Google Calendar, and make to-do lists with Google Duties.
Google says it’s updating Deep Analysis, Gemini’s AI agent that generates thorough analysis stories, by permitting customers to add their very own non-public PDFs and pictures.
Sew
Sew is an AI-powered instrument to assist individuals design net and cell app entrance ends by producing the mandatory UI parts and code. Sew may be prompted to create app UIs with just a few phrases and even a picture, offering HTML and CSS markup for the designs it generates.
Sew is a little more restricted in what it may well do in comparison with another vibe coding merchandise, however there’s a good quantity of customization choices.
Google has additionally expanded entry to Jules, its AI agent geared toward serving to builders repair bugs in code. The instrument helps builders perceive complicated code, create pull requests on GitHub, and deal with sure backlog objects and programming duties.
Mission Mariner
Mission Mariner is Google’s experimental AI agent that browses and makes use of web sites. Google says it has considerably up to date how Mission Mariner works, permitting the agent to tackle practically a dozen duties at a time, and is now rolling it out to customers.
For instance, Mission Mariner customers can buy tickets to a baseball recreation or purchase groceries on-line with out ever visiting a third-party web site. Folks can simply chat with Google’s AI agent, and it visits web sites and takes actions for them.
Mission Astra
Google’s low latency, multimodal AI expertise, Mission Astra, will energy an array of recent experiences in Search, the Gemini AI app, and merchandise from third-party builders.
Mission Astra was born out of Google DeepMind as a option to showcase practically real-time, multimodal AI capabilities. The corporate says it’s now constructing these Mission Astra glasses with companions together with Samsung and Warby Parker, however the firm doesn’t have a set launch date but.
AI Mode
Google is rolling out AI Mode, the experimental Google Search characteristic that lets individuals ask complicated, multi-part questions through an AI interface, to customers within the U.S. this week.
AI Mode will help using complicated knowledge in sports activities and finance queries, and it’ll provide “try it on” choices for attire. Search Reside, which is rolling out later this summer season, will allow you to ask questions primarily based on what your telephone’s digital camera is seeing in real-time.
Gmail is the primary app to be supported with customized context.
Beam 3D teleconferencing
Beam, beforehand referred to as Starline, makes use of a mix of software program and {hardware}, together with a six-camera array and customized mild subject show, to let a person converse with somebody as in the event that they have been in the identical assembly room. An AI mannequin converts video from the cameras, that are positioned at completely different angles and pointed towards the person, right into a 3D rendering.
Google’s Beam boasts “near-perfect” millimeter-level head monitoring and 60fps video streaming. When used with Google Meet, Beam supplies an AI-powered real-time speech translation characteristic that preserves the unique speaker’s voice, tone, and expressions.
And talking of Google Meet, Google introduced that Meet is getting real-time speech translation.
Extra AI updates
Google is launching Gemini in Chrome, which is able to give individuals entry to a brand new AI shopping assistant that can assist them shortly perceive the context of a web page and get duties carried out.
Gemma 3n is a mannequin designed to run “smoothly” on telephones, laptops, and tablets. It’s obtainable in preview beginning Tuesday; it may well deal with audio, textual content, photos, and movies, in line with Google.
The corporate additionally introduced a ton of AI Workspace options coming to Gmail, Google Docs, and Google Vids. Most notably, Gmail is getting customized good replies and a brand new inbox-cleaning characteristic, whereas Vids is getting new methods to create and edit content material.
Video Overviews are coming to NotebookLM, and the corporate rolled out SynthID Detector, a verification portal that makes use of Google’s SynthID watermarking expertise to assist determine AI-generated content material. Lyria RealTime, the AI mannequin that powers its experimental music manufacturing app, is now obtainable through an API.
Put on OS 6
Put on OS 6 brings a unified font to tiles for a cleaner app look, and Pixel Watches are getting dynamic theming that syncs app colours with watch faces.
The core promise of the brand new design reference platform is to let builders construct higher customization in apps together with seamless transitions. The corporate is releasing a design guideline for builders together with Figma design recordsdata.
Google Play
Google is beefing up the Play Retailer for Android builders with contemporary instruments to deal with subscriptions, matter pages so customers can dive into particular pursuits, audio samples to present of us a sneak peek into app content material, and a brand new checkout expertise to make promoting add-ons smoother.
“Topic browse” pages for films and reveals (U.S. just for now) will join customers to apps tied to tons of reveals and flicks. Plus, builders are getting devoted pages for testing and releases, and instruments to regulate and enhance their app rollouts. Builders utilizing Google also can now halt reside app releases if a vital drawback pops up.
Subscription administration instruments are additionally getting an improve with multi-product checkout. Devs will quickly have the ability to provide subscription add-ons alongside most important subscriptions, all below one fee.
Android Studio
Android Studio is integrating new AI options, together with “Journeys,” an “agentic AI” functionality that coincides with the discharge of the Gemini 2.5 Professional mannequin. And an “Agent Mode” will have the ability to deal with more-intricate improvement processes.
Android Studio will obtain new AI capabilities, together with an enhanced “crash insights” characteristic within the App High quality Insights panel. This enchancment, powered by Gemini, will analyze an app’s supply code to determine potential causes of crashes and recommend fixes.
