Google I/O 2025: Top 100 Announcements

That’s a wrap on I/O 2025! Here’s what was announced, launched and demoed. Yesterday at Google I/O, it was shared how progress in AI is being applied across products. Major upgrades are coming to the Gemini app, generative AI tools and everything in between — including progress being made with AI models and new ways to access them. Here’s a list of I/O 2025’s highlights.

AI Updates

AI Mode is starting to roll out for everyone in the U.S. right on Search. For access right away, opt in via Labs.
Deep research capabilities are coming to AI Mode in Labs, with Deep Search, for questions where you want an even more thorough response.
Live capabilities from Project Astra are coming to AI Mode in Labs. With Search Live, available this summer, you can talk back-and-forth with Search about what you see in real-time, using your camera.
Agentic capabilities from Project Mariner are coming to AI Mode in Labs, starting with event tickets, restaurant reservations and local appointments.
AI Mode in Labs will analyze complex datasets and create graphics that bring them to life, custom built for your query, when you need some extra help crunching numbers or visualizing data. This will be available for sports and finance queries.
A new AI Mode shopping experience is being introduced that brings together advanced AI capabilities with the Shopping Graph to help you browse for inspiration, think through considerations and find the right product.

Search and Shopping

You can virtually try on billions of apparel listings just by uploading a photo of yourself. The “try on” experiment is rolling out to Search Labs users in the U.S.
A new agentic checkout was also shown to help you buy at a price that fits your budget with ease. Just tap “track price” on any product listing, set what you want to spend and you’ll be notified if the price drops.
Updates on AI Overviews were shared: Since last year’s I/O, AI Overviews have scaled up to 1.5 billion monthly users in 200 countries and territories. Google Search is bringing generative AI to more people than any other product in the world.
In markets like the U.S. and India, AI Overviews is driving over 10% increase in usage of Google for the types of queries that show AI Overviews.
Gemini 2.5 is coming to Search for both AI Mode and AI Overviews in the U.S. this week.

Gemini Updates

Gemini is now an even better study partner with a new interactive quiz feature. Simply ask Gemini to “create a practice quiz on…” and Gemini will generate questions.
In the coming weeks Gemini Live will also be made more personal by connecting some of your favorite Google apps so you can take actions mid-conversation, like adding something to your calendar or asking for more details about a location. This will start with Google Maps, Calendar, Tasks and Keep, with more app connections coming later.
Camera and screen sharing capabilities for Gemini Live are beginning to roll out beyond Android to Gemini app users on iOS.
A new Create menu within Canvas is being introduced that helps you explore the breadth of what Canvas can build for you, allowing you to transform text into interactive infographics, web pages, immersive quizzes and even podcast-style Audio Overviews in 45 languages.
You can upload PDFs and images directly into Deep Research so your research reports draw from a combination of public information and details that you provide.
Soon, you’ll be able to link your documents from Drive or from Gmail and customize the sources Deep Research pulls from, like academic literature.
Agent Mode was announced, an experimental feature where you will be able to simply describe your end goal and Gemini can get things done on your behalf. An experimental version of Agent Mode in the Gemini app will be coming soon to Google AI Ultra subscribers.
Gemini in Chrome will begin rolling out on desktop to Google AI Pro and Google AI Ultra subscribers in the U.S. who use English as their Chrome language on Windows and macOS.
The Gemini app now has over 400 million monthly active users.
With the latest update, Gemini 2.5 Pro is now the world-leading model across the WebDev Arena and LMArena leaderboards.
LearnLM is being infused directly into Gemini 2.5, which is now the world’s leading model for learning. Gemini 2.5 Pro outperformed competitors on every category of learning science principles.
A new preview version of the leading model, Gemini 2.5 Flash, was introduced, with stronger performance on coding and complex reasoning tasks that is optimized for speed and efficiency.
2.5 Flash is now available to everyone in the Gemini app, and an updated version will be made generally available in Google AI Studio for developers and in Vertex AI for enterprises in early June, with 2.5 Pro soon after.
2.5 Pro will get even better with Deep Think, an experimental, enhanced reasoning mode for highly-complex math and coding.
New capabilities are being brought to both 2.5 Pro and 2.5 Flash, including advanced security safeguards. A new security approach helped significantly increase Gemini’s protection rate against indirect prompt injection attacks during tool use, making Gemini 2.5 the most secure model family to date.
Project Mariner's computer use capabilities are being brought into the Gemini API and Vertex AI. Companies like Automation Anywhere, UiPath, Browserbase, Autotab, The Interaction Company and Cartwheel are exploring its potential, and it will be rolled out more broadly for developers to experiment with this summer.
Both 2.5 Pro and Flash will now include thought summaries in the Gemini API and in Vertex AI. Thought summaries take the model’s raw thoughts and organize them into a clear format with headers, key details and information about model actions, like when they use tools.
2.5 Flash was launched with thinking budgets to give developers more control over cost by balancing latency and quality, and this capability is being extended to 2.5 Pro. This allows you to control the number of tokens a model uses to think before it responds, or even turn its thinking capabilities off. Gemini 2.5 Pro with budgets will be generally available for stable production use in the coming weeks, along with the generally available model.
Native SDK support for Model Context Protocol (MCP) definitions was added in the Gemini API for easier integration with open-source tools. Ways to deploy MCP servers and other hosted tools are also being explored, making it easier for you to build agentic applications.
A new research model, called Gemini Diffusion, was introduced. This text diffusion model learns to generate outputs by converting random noise into coherent text or code, like how the current models in image and video generation work. Work will continue on different approaches to lowering latency in all Gemini models, with a faster 2.5 Flash Lite coming soon.

New AI Subscription Plans

Google AI Ultra, a new AI subscription plan with the highest usage limits and access to the most capable models and premium features, plus 30 TB of storage and access to YouTube Premium, was introduced.
Google AI Ultra is available in the U.S. now, with more countries coming soon. It’s $249.99 a month, with a special offer for first-time users of 50% off for your first three months.
College students in the U.S., Brazil, Indonesia, Japan and the U.K. are also eligible to get a free upgrade of Gemini for a whole school year — more countries are coming soon.
There’s also Google Al Pro, which gives you a suite of Al tools for $19.99/month. This Pro plan will level up your Gemini app experience. It also includes products like Flow, NotebookLM and more, all with special features and higher rate limits.

Veo and Imagen Updates

Veo 3, which lets you generate video with audio and is now available in the Gemini app for Google AI Ultra subscribers in the U.S., as well as in Vertex AI, was announced.
New capabilities were also added to the popular Veo 2 model, including new camera controls, outpainting and object add and remove.
Four new films created with Veo alongside other tools and techniques were shown. View these films from partners and other inspirational content on Flow TV.
Imagen 4 is the latest Imagen model, and it has clarity in fine details like skin, fur and intricate textures, and excels in both photorealistic and abstract styles. Imagen 4 is available in the Gemini app.
Imagen 4 is also available in Whisk, and to enterprises in Vertex AI.
Soon, Imagen 4 will be available in a Fast version that’s up to 10x faster than Imagen 3.
Imagen 4 can create images in a range of aspect ratios and up to 2K resolution so you can get even higher-quality for printing and presentations.
It is also significantly better at spelling and typography, making it easier to create your own greeting cards, posters and even comics.
Flow is a new AI filmmaking tool. Using Google DeepMind’s models, Flow lets you weave cinematic films with control of characters, scenes and styles, so more people than ever can create visually striking movies with AI.
Flow is available for Google AI Pro and Ultra plan subscribers in the United States.

Music AI Sandbox and DeepMind Partnership

In April, access was expanded to Music AI Sandbox, powered by Lyria 2. Lyria 2 brings composition and exploration, and is now available for creators through YouTube Shorts and enterprises in Vertex AI.
Lyria 2 can arrange rich vocals that sound like a solo singer or a full choir.
Lyria RealTime is an interactive music generation model that allows anyone to interactively create, control, and perform music in real time. This model is now available via the Gemini API in Google AI Studio and Vertex AI.
A partnership between Google DeepMind and Primordial Soup, a new venture dedicated to storytelling innovation founded by director Darren Aronofsky, was announced. Primordial Soup is producing three short films using Google DeepMind’s generative AI models, tools and capabilities, including Veo.
The first film, “ANCESTRA,” is directed by filmmaker Eliza McNitt and will premiere at the Tribeca Festival on June 13, 2025.

SynthID Detector

To make it easier for people and organizations to detect AI-generated content, SynthID Detector, a verification portal that helps to quickly and efficiently identify content that is watermarked with SynthID, was announced.
Since launch, SynthID has already watermarked over 10 billion pieces of content.
The SynthID Detector portal is starting to roll out to a group of early testers. Journalists, media professionals and researchers can join a waitlist to gain access to the SynthID Detector.

Project Astra Updates

Work is being done to extend Gemini 2.5 Pro to become a “world model” that can make plans and imagine new experiences by understanding and simulating aspects of the world.
Updates to Project Astra, a research prototype that explores the capabilities of a universal AI assistant, include more natural voice output with native audio, improved memory and computer control. Over time these new capabilities will be brought to Gemini Live and new experiences in Search, Live API for devs and new form factors like Android XR glasses.
As part of Project Astra research, a partnership was made with the visual interpreting service Aira to build a prototype that assists members of the blind and low-vision community with everyday tasks, complementing the skills and tools they already use.
With Project Astra, a conversational tutor is being prototyped that can help with homework. Not only can it follow along with what you’re working on, but it can also walk you through problems step-by-step, identify mistakes and even generate diagrams to help explain concepts if you get stuck.
This research experience will be coming to Google products later this year and Android Trusted Testers can sign up for a waitlist for a preview.

Android XR and Google Beam

The first Android XR device coming later this year, Samsung’s Project Moohan, was previewed. This headset will offer immersive experiences on an infinite screen.
A sneak peek was shared at how Gemini will work on glasses with Android XR in real-world scenarios, including messaging friends, making appointments, asking for turn-by-turn directions, taking photos and more.
Live language translation between two people was demoed, showing the potential for these glasses to break down language barriers.
Android XR prototype glasses are now in the hands of trusted testers, who are helping make sure a product is being built that respects privacy.
Partnerships with eyewear brands, starting with Gentle Monster and Warby Parker, are being made to create glasses with Android XR that you’ll want to wear all day.
The partnership with Samsung is advancing to go beyond headsets and extend Android XR to glasses. Together a software and reference hardware platform is being created that will enable the ecosystem to make great glasses. Developers will be able to start building for this platform later this year.
Project Starline, a research project that enabled remote conversations that used 3D video technology to make it feel like two people were in the same room, is evolving into a new platform called Google Beam.
Work with Zoom and HP is being done to bring the first Google Beam devices to market with select customers later this year. Partnerships with industry leaders like Zoom, Diversified and AVI-SPL are also being made to bring Google Beam to businesses and organizations worldwide.
The first Google Beam products from HP will be at InfoComm in a few weeks.

Google Meet and Developer Updates

Speech translation, which is available now in Google Meet, was announced. This translation feature happens in near real-time, and is able to maintain the quality, tone, and expressiveness of someone’s voice. The conversation enables people to understand each other and feel connected, with no language barrier.
Over 7 million developers are building with Gemini, five times more than this time last year.
Gemini usage on Vertex AI is up 40 times compared to this time last year.
New previews are being released for text-to-speech in 2.5 Pro and 2.5 Flash. These have support for multiple speakers, enabling text-to-speech with two voices via native audio out. Like Native Audio dialogue, text-to-speech is expressive, and can capture nuances, such as whispers. It works in over 24 languages and seamlessly switches between them.
The Live API is introducing a preview version of audio-visual input and native audio out dialogue, so you can directly build conversational experiences.
Jules is a parallel, asynchronous agent for GitHub repositories to help you improve and understand your codebase. It is now open to all developers in beta. With Jules you can delegate multiple backlog items and coding tasks at the same time, and even get an audio overview of all the recent updates to your codebase.
Gemma 3n is the latest fast and efficient open multimodal model that’s engineered to run smoothly on your phones, laptops, and tablets. It handles audio, text, image, and video. The initial rollout is underway on Google AI Studio and Google Cloud with plans to expand to open-source tools in the coming weeks.
Google AI Studio now has a cleaner UI, integrated documentation, usage dashboards, new apps, and a new Generate Media tab to explore and experiment with cutting-edge generative models, including Imagen, Veo and native image generation.
Colab will soon be a new, fully agentic experience. Simply tell Colab what you want to achieve, and watch as it takes action in your notebook, fixing errors and transforming code to help you solve hard problems faster.
SignGemma is an upcoming open model that translates sign language into spoken language text, (best at American Sign Language to English), enabling developers to create new apps and integrations for Deaf and Hard of Hearing users.
MedGemma is the most capable open model for multimodal medical text and image comprehension designed for developers to adapt and build their health applications, like analyzing medical images. MedGemma is available now for use now as part of Health AI Developer Foundations.
Stitch is a new AI-powered tool to generate UI designs and corresponding frontend code for desktop and mobile by using natural language descriptions or image prompts.
Journeys in Android Studio lets developers test critical user journeys using Gemini by describing test steps in natural language.
Version Upgrade Agent in Android Studio is coming soon to automatically update dependencies to the latest compatible version, parsing through release notes, building the project and fixing any errors.
New updates across the Google Pay API are designed to help developers create checkout experiences, including Google Pay in Android WebViews.
Flutter 3.32 has new features designed to accelerate development and enhance apps.
Updates for the Agent Development Kit (ADK), the Vertex AI Agent Engine, and the Agent2Agent (A2A) protocol, which enables interactions between multiple agents, were shared.
Developer Preview for Wear OS 6 introduces Material 3 Expressive and updated developer tools for Watch Faces, richer media controls and the Credential Manager for authentication.
Gemini Code Assist for individuals and Gemini Code Assist for GitHub are generally available, and developers can get started in less than a minute. Gemini 2.5 now powers both the free and paid versions of Gemini Code Assist, features advanced coding performance; and helps developers excel at tasks like creating visually compelling web apps, along with code transformation and editing.
Here’s an example of a recent update you can explore in Gemini Code Assist: Quickly resume where you left off and jump into new directions with chat history and threads.
Firebase announced new features and tools to help developers build AI-powered apps more easily, including updates to the recently launched Firebase Studio and Firebase AI Logic, which enables developers to integrate AI into their apps faster.
A new Google Cloud and NVIDIA developer community, a dedicated forum to connect with experts from both companies, was introduced.
Google AI Edge Portal in private preview, a new solution from Google Cloud for testing and benchmarking on-device machine learning (ML) at scale, was introduced.

Gmail, Google Vids, and NotebookLM Updates

Gmail is getting new, personalized smart replies that incorporate your own context and tone. They’ll pull from your past emails and files in your Drive to draft a response, while also matching your typical tone so your replies sound like you. Try it yourself later this year.
Google Vids is now available to Google AI Pro and Ultra users.
The NotebookLM app is now available on Play Store and App Store, to help users take Audio Overviews on the go.
For NotebookLM, more flexibility is being brought to Audio Overviews, allowing you to select the ideal length for your summaries, whether you prefer a quick overview or a deeper exploration.
Video Overviews are coming soon to NotebookLM, helping you turn information like PDFs, docs, images, diagrams and key quotes into narrated overviews.
A NotebookLM notebook was shared which included previews of Video Overviews!
Sparkify helps you turn your questions into a short animated video, made possible by the latest Gemini and Veo models. These capabilities will be coming to Google products later this year, but you can sign up for a waitlist for a chance to try it out.
Improvements based on feedback are also being brought to Learn About, an experiment in Labs where conversational AI meets your curiosity.
People are adopting AI more than ever before. This time last year, 9.7 trillion tokens a month were being processed across products and APIs. Now, over 480 trillion are being processed — 50 times more.