Google held its annual I/O developer conference at the Shoreline Amphitheatre in Mountain View on May 19, 2026. Over a packed two-hour keynote, the company laid out a vision where artificial intelligence moves from answering questions to taking action across every product. Here is a straightforward breakdown of everything announced, without the fluff.

Image Credit: Google

Gemini 3.5 Flash

Google launched Gemini 3.5 Flash, a model built for speed and long-running tasks. It runs four times faster than other frontier models. On a coding benchmark called Terminal-Bench 2.1, it scored 76.2 percent, beating its predecessor Gemini 3.1 Pro. It also posted a 1656 Elo score on a real-world economic task test.

The model can build apps, maintain code, prepare financial documents, and migrate old codebases to modern frameworks. It generated six payment page designs in under a minute during a live demo. Gemini 3.5 Flash is now available in the Gemini app, Google AI Studio, Android Studio, and the Antigravity developer platform. A more powerful Gemini 3.5 Pro model is already in internal use and will launch in June.

Gemini Omni: Video Generation from Any Input

Gemini Omni is a new creative model that can generate video from text, photos, audio, or existing video clips. It understands real-world physics like gravity and fluid dynamics, which makes the generated movements more realistic. Every frame carries an invisible SynthID watermark for transparency.

The first version, Omni Flash, is available inside the Gemini app, Google Flow, and YouTube Shorts. A demo showed it generating a video of a violinist, then changing the background, making the violin transparent, and shifting the camera angle, all through spoken instructions.

Gemini Spark: A 24-Hour Personal AI Agent

Gemini Spark is a persistent AI agent that runs on dedicated virtual machines inside Google Cloud. It stays active even when your devices are off. Spark connects to Gmail, Docs, Sheets, Slides, and third-party apps like Canva and Instacart through the Model Context Protocol.

It can automatically flag new subscription charges from credit card statements, monitor emails, draft replies, create study guides, and manage workflows across multiple apps. A test group gets access the week of the conference. A beta for Google AI Ultra subscribers in the United States follows the next week. Users will soon be able to interact with Spark through email and chat.

Daily Brief: A Personalised Morning Summary

A new Daily Brief feature in the Gemini app pulls information from Gmail, Calendar, and connected services to present a morning digest. It highlights the most important actions and prepares you for the day ahead.

Gemini App Redesigned

The Gemini app received a visual overhaul called Neural Expressive. It brings fluid animations, a new colour palette, a custom typeface, and haptic feedback. The live voice chat now supports multiple regional accents in several languages. The update is designed to make the app feel more natural and responsive.

Antigravity 2.0

Antigravity 2.0 is a standalone desktop app for building and managing multiple AI agents at once. One agent can create brand assets while another codes a website simultaneously. It includes a terminal sandbox for safe testing, credential masking, and reinforced Git strategies. A new command-line tool called Antigravity CLI and an Android CLI that lets agents directly call Android Studio functions were also launched. Google AI Studio now supports vibe-coding entire native Android apps and publishing them straight to the Play Store.

Google Search

The traditional search box is being replaced by an intelligent search box that accepts text, images, files, screenshots, and open Chrome tabs. A query like "What is a black hole" can return an AI-generated explanatory video directly in the results. A new Generative UI creates different layouts for different answer types, generated in real time. The intelligent search box started rolling out immediately. The full Generative UI experience arrives over the summer.

Universal Cart: Cross-Web Shopping Agent

Universal Cart is an AI shopping system that works across the entire web. You describe what you need, and the agent searches, compares prices, checks availability, and builds a cart across multiple retailers. It connects with Ask Maps, letting you plan a trip and shop for everything needed in one flow.

Intelligent Eyewear: Android XR Glasses

Google announced a new line of Intelligent Eyewear running Android XR. The first product is an audio-focused pair of glasses launching in autumn. It delivers private audio from Gemini into the wearer's ear, can take photos, make calls, and interact with apps without a screen. More advanced Android XR glasses with visual displays were shown in prototype form under Project Aura.

Googlebooks

At a pre-I/O event called The Android Show, Google introduced Googlebooks, premium laptops that will eventually replace Chromebooks. They run a unified operating system built from Android 17 and ChromeOS, codenamed Aluminium OS. The hardware features a glowing keyboard strip called the Glowbar. The software highlight is Magic Pointer, which uses Gemini Intelligence to surface contextual actions when you hover over anything on screen, like creating a calendar event from a date in an email. Acer, ASUS, Dell, HP, and Lenovo are manufacturing partners. First models ship in autumn.

Android 17: Gemini Intelligence Embedded

Android 17, now in public beta, embeds Gemini Intelligence directly into the operating system. It can find a syllabus in Gmail, identify required textbooks, and add them to a shopping cart without switching apps. Smart Autofill handles forms across apps and Chrome. Rambler, a speech-to-text tool in Gboard, cleans filler words and restructures dictated sentences. Create My Widget lets you describe a widget in plain language and generates it instantly. All Android emojis are being redesigned in 3D under the name Noto 3D. A new Pause Point feature adds a ten-second pause when opening distracting apps. Android Auto now extends Gemini Intelligence into the car when a paired phone supports it.

Pixel 11 Hint

Google did not formally announce the Pixel 11, but a brief shot during the Omni demo showed a phone with a thin light strip around the rear camera bar. This matches reports of Pixel Glow, an LED notification system using up to eight colours to flag important alerts. The Pixel 11 series is expected to launch at a separate Made by Google event, likely in August, powered by the Tensor G6 chip built on a two-nanometer process.

Developer Tools and Infrastructure

The Gemini API now processes roughly 190 billion tokens per minute. Over 375 Google Cloud customers have each processed more than one trillion tokens in the past year. New agentic AI toolkits for Python and TypeScript were released. A Managed Agents API allows building and running custom agents in secure, Google-hosted environments. CodeMender is a new security agent that finds and fixes code vulnerabilities automatically. WebMCP, a new open standard in Chrome 149, lets websites expose tools that AI agents can use. Chrome DevTools is now open to AI agents for automated debugging.

AI Subscription Plans Restructured

A new AI Ultra tier costs one hundred dollars per month, offering five times the usage of the Pro plan, priority access to Antigravity, and twenty terabytes of storage. The previous top-tier plan dropped from two hundred and fifty dollars to two hundred dollars. Google is also shifting from daily prompt count limits to a compute consumption model that refreshes every five hours up to a weekly cap.

Gemini for Science and a Bold Statement

DeepMind CEO Demis Hassabis introduced Gemini for Science, a set of AI tools built to accelerate scientific research. He described the current moment as standing in the foothills of the singularity and framed the technology as a force multiplier for human ingenuity. The statement captured the scale of ambition behind every announcement made on stage.

Some data may be wrong. The information is based on general researches in internet. Weniba doesn't guarantee for correctness. Weniba doesn't take responsibility for wrong information.