Google I/O 2025: From Gemini to Astra- Know All Updates & Announcements In Sundar Pichai's Keynote

The keynote for Google I/O 2025 will commence at 10:30 pm IST on May 20 at the Shoreline Amphitheatre in Mountain View. You can watch Google's keynote live on YouTube.

At 2:00 am Wednesday (IST), the developer keynote will begin. As the name suggests, the developer keynote will focus less on consumer-oriented topics.

While Gemini AI’s rollout to Wear OS and Android Auto continues, Google I/O 2025 is expected to focus heavily on extended reality.

Updates on Project Moohan—the mixed reality headset built with Samsung—will likely headline the event, showcasing features like real-time object recognition, 3D mapping, and seamless Gemini AI integration.

Google may also preview new XR demos and smart glasses concepts, highlighting Android XR’s evolving capabilities.

Ahead of the Google I/O 2025 keynote, Toro y Moi delivered a live performance that doubled as an AI experiment.

Jamming alongside AI tools potentially Gemini or Lyria the artist explored real-time collaboration between human and machine, fusing chillwave sounds with generative tech. The set teased Google’s AI creative ambitions while setting an immersive, forward-thinking vibe for the event.

As the crowd settled in at Shoreline Amphitheatre, Google employees opened Google I/O 2025 with a live coding session, building a to-do list app onstage.

The demo, likely powered by tools like Android Studio, Flutter, and Gemini AI integration, set a collaborative, AI-centric vibe for the day — spotlighting how AI is transforming developer workflows from the first lines of code.

During his Google I/O 2025 keynote, Sundar Pichai spotlighted Gemini 2.5 Pro’s rise to the top of LMArena’s Chatbot Arena, outperforming rivals like Anthropic’s Claude in categories including Coding, Creative Writing, and Vision.

Pichai also noted an 80% usage spike for Gemini 2.5 Pro in April on Google AI Studio and the Gemini API, calling it the fastest-growing AI model on coding platforms like Cursor.

At Google I/O 2025, Gemini 2.5 Pro earned viral attention for completing the 1996 Game Boy classic Pokémon Blue. CEO Sundar Pichai marked the achievement on X with a playful nod to “Artificial Pokémon Intelligence.”

The milestone, streamed via Joel Z’s independent “Gemini Plays Pokémon” project, showcased an agentic AI harness executing over 106,000 in-game actions to beat the title a moment of both nostalgia and AI prowess.

CEO Sundar Pichai highlighted at Google I/O 2025 that Gemini 2.5 Pro achieved a clean sweep across all categories in LMArena’s Chatbot Arena benchmarks, outperforming competitors in coding, creative writing, vision, and more.

He also noted an 80% usage surge in April—making Gemini the fastest-growing AI model on developer platform Cursor—underscoring its rapid adoption and versatility.

At Google I/O 2025, Sundar Pichai announced that the Gemini app has hit 400 million monthly active users, marking a major milestone in the AI platform’s rapid global adoption.

At Google I/O 2025, Google introduced Beam, a new AI-driven communications service designed to enhance real-time messaging, collaboration, and agentic interactions across devices and apps.

Google announced a collaboration with HP to integrate its new Beam AI communications service into HP devices, with availability slated for the end of the year.

This partnership aims to deliver seamless, AI-powered messaging and collaborative features directly on HP hardware, expanding Beam’s reach across enterprise and consumer markets.

Google has launched a new live translation feature in Google Meet—powered by its Gemini AI—that translates spoken language into the listener’s preferred language while preserving the speaker’s original voice, tone, and inflection.

Starting today, Gemini Live, Google’s AI-powered conversational assistant is getting camera and screen sharing support on both Android and iOS.

This means users can show Gemini what they’re seeing in real time or share their screen for contextual help, troubleshooting, or on-the-fly information gathering. It’s another big step in making AI assistants more interactive and hands-on in everyday tasks.

Announced at last year’s Google I/O, Project Mariner, the initiative enabling Gemini to autonomously perform tasks on a user’s computer is officially headed to the Gemini API. Soon, developers will be able to tap into this functionality, allowing AI agents to proactively manage tasks without requiring user prompts, marking a major leap in ambient, agentic computing.

Google announced that Agent Mode, a feature enabling Gemini to act more autonomously on behalf of users, will soon be available to Gemini subscribers.

This mode builds on Google’s growing emphasis on agentic AI, allowing Gemini to handle tasks, manage workflows, and interact with apps with minimal input — hinting at a future of AI-powered personal assistants with deeper system access and initiative.

At Google I/O 2025, Google confirmed that Personalized Smart Replies are coming to Gmail this summer, initially for Gemini subscribers.

The AI-powered feature will tailor suggested email replies based on each user’s writing style, habits, and context. Many of Google’s latest AI upgrades appear to be debuting under the Gemini subscription tier before broader release.

At Google I/O 2025, Google introduced Gemini 2.5 Flash, a faster, more efficient sibling to Gemini 2.5 Pro claiming improvements “in nearly every dimension.” The lightweight model will roll out in June, with early access via Google AI Studio, Vertex AI, and the Gemini app.

Google announced at I/O 2025 that new text-to-speech features powered by Gemini 2.5 Flash and Pro are now live in Gemini. The system supports native audio output in 24 languages and can fluidly switch between them maintaining the same voice across transitions.

At Google I/O 2025, Google revealed Gemini 2.5 Flash is now 22% more efficient, delivering the same quality responses while using fewer tokens.

A new “Thinking Budgets” option will also let users control how many tokens Gemini can use before replying giving developers and users more flexibility over AI response behavior.

At Google I/O 2025, Google announced that Gemini 2.5 Pro’s multimodal abilities now allow developers to update code using hand-drawn sketches, bypassing the need for complex APIs or programming languages.

This feature is coming to various AI coding tools and Google’s own developer products, streamlining workflows and lowering the barrier to entry for coding.

At Google I/O 2025, Google unveiled Gemini Diffusion, a blazing-fast AI model capable of generating code five times quicker than its lightest Gemini 2.5 variant. The demo showed it parsing and solving coding prompts almost instantly so fast, if you blinked, you missed it.

At Google I/O 2025, Google announced that Project Astra’s advanced AI capabilities will be coming to Gemini Live.

This integration promises smarter, faster, and more context-aware real-time interactions on both Android and iOS, leveling up how users engage with AI through their devices.

AlphaFold, an AI language model developed to predict protein structures, is accelerating medical research by providing critical insights into molecular biology.

Its breakthroughs help scientists better understand diseases and develop new treatments, transforming the pace and precision of biomedical discovery.

AI Mode brings Gemini-powered chat capabilities directly into Google Search, allowing users to ask complex questions and engage in follow-ups seamlessly from the search bar.

This new feature transforms traditional search into an interactive, conversational experience, making information access faster and more intuitive.

After holding back Play Store news during its Android pre-show, Google revealed a suite of marketing-focused enhancements to help developers better showcase their apps and services.

The upgrades include improved promotional tools, richer media support, and advanced analytics to optimize store listings and reach targeted audiences—enabling developers to craft more engaging storefronts and drive discovery directly within the Play Store.

Google announced that Project Mariner’s powerful capabilities will be integrated into AI Mode on Search this summer.

In a demo, AI Mode swiftly found and reserved baseball game tickets, taking users directly to checkout within moments—showcasing how AI can handle complex, multi-step tasks effortlessly and improve user experience.

Google’s new Search Live feature lets your camera act as an assistant by analyzing your surroundings and delivering instant answers to your questions in real time, enhancing the way you interact with the world through visual search.

Soon, Google Search will let you monitor product prices, notify you when they hit your desired target, and allow seamless one-tap purchases—all rolling out later in 2025 for a smoother shopping experience.

At Google I/O 2025, Google announced that Gemini will soon incorporate richer personal context pulled from across Google apps, enabling more personalized and relevant AI interactions.

Google Search now offers a virtual try-on experience, helping users see how clothes will fit before purchasing, enhancing online shopping with AI-powered visualization.

At Google I/O 2025, it was announced that Gemini is being integrated into Chrome for Windows and macOS, with the rollout starting today, May 20, 2025, in the United States.

Users can access Gemini via a sparkle icon next to the window controls, keyboard shortcuts (Ctrl+G on Mac, Alt+G on Windows), or through the taskbar. You can type queries or use Gemini Live for voice interaction.

At Google I/O 2025, Google introduced Imagen 4, the latest iteration of its text-to-image diffusion model.

Imagen 4 delivers faster inference and lower compute requirements, while offering enhanced creative control—producing more detailed, stylistically diverse images.

Google announced Veo 3, its latest AI-driven video generation model, now available.

Veo 3 can create fully produced videos—complete with sound effects, dialogue, and character voices—directly from text prompts, marking a major step forward in end-to-end multimedia generation.

Google announced that all Gemini Live features are now free for users, enabling immediate access to advanced AI tools like Veo 3 for video generation and Imagen 4 for image creation, along with multimodal interactions and real-time assistance—no subscription required.

Indian playback singer and composer Shankar Mahadevan lauded Google’s AI Sandbox for its intuitive, musician-friendly features. In sharing his firsthand experience, he highlighted how the platform’s generative tools helped him experiment with melodies, harmonies, and production ideas—streamlining creative workflows and inspiring new musical directions.

Google has launched two new AI service tiers—Google AI Pro and Google AI Ultra—designed to offer progressively advanced features and performance for power users. Available in the U.S. and rolling out soon, these subscription plans aim to provide enhanced access to Google’s suite of AI tools across Search, Workspace, and the Gemini ecosystem.

During a live demo at Google I/O, Nishtha Bhatia used Android XR–powered glasses to livestream while interacting with Gemini. As she walked backstage, she asked Gemini to recall the name and details of the coffee shop printed on her mug, and the AI instantly displayed both the name and a description—all in real time during the stream.

Google announced its new Android XR smartglasses—wearable AR eyewear that lets users send texts, mute notifications, record live video, and more hands-free. Designed for seamless integration with Android devices, the glasses aim to enhance everyday tasks with intuitive, on-the-go controls and immersive XR experiences.

Google announced a collaboration with Samsung to co-develop next-generation XR headsets and smartglasses, combining Samsung’s hardware expertise with Google’s Android XR platform and AI capabilities.

The partnership aims to accelerate the rollout of immersive wearable devices, offering features like hands-free messaging, real-time translation, and context-aware AI assistance.