Google Debuts Gemini and Android Updates at I/O 2026

Google held its I/O 2026 developer keynote on May 19, 2026, led by CEO Sundar Pichai, with Google DeepMind CEO Demis Hassabis presenting the model updates. The company announced Gemini 3.5 Flash, which it says outperforms the previous Gemini 3.1 Pro on coding and agentic benchmarks while running roughly four times faster than competing frontier models; it is now available in the Gemini app, Google Search's AI Mode, Google AI Studio, and Android Studio. Google also unveiled Gemini Omni (initial version "Omni Flash"), a tool that generates video from text, photo, video, or audio input, rolling out the same day to Google AI Plus, Pro, and Ultra subscribers as well as YouTube Shorts and YouTube Create users.

For developers, Google upgraded Antigravity, its agent-development platform, to version 2.0 with a standalone desktop app, a new command-line interface, and built-in terminal sandboxing, credential masking, and Git-policy hardening for running subagents. New Google AI Studio integrations add native Kotlin support for building Android apps and one-click deployment to Cloud Run, and a new Managed Agents API removes infrastructure setup for hosted agents. Google also introduced WebMCP, a proposed open web standard (origin trial starting in Chrome 149) that lets developers expose page functions and forms so browser-based AI agents can interact with sites directly, plus a stable Android CLI and open-sourced "Android skills" that let any LLM or coding agent perform tasks like Jetpack Compose migrations.

On the consumer side, Google introduced Gemini Spark, described as a personal AI agent that runs continuously in the background on a user's behalf, rolling out in beta to US Google AI Ultra subscribers the following week before expanding to Gmail, Workspace, Android, and iOS. A companion feature, Android Halo, will show users a persistent on-screen indicator of what their agents are doing; it is slated to arrive later this year alongside Spark. Google Search picked up an AI-powered, dynamically expanding search box built on Gemini 3.5 Flash, user-buildable "information agents" that continuously monitor for things like apartment listings, and a Universal Cart feature paired with a new Agent Payments Protocol (AP2), which Google says lets users set spending guardrails for agent-initiated purchases. Pricing changed too: a new $100/month AI Ultra tier launched for developers and power users (five times the Gemini app usage limits of the Pro plan, plus Antigravity priority access and 20TB of storage), while Google cut its existing top-tier Ultra plan from $250 to $200/month.

Notably, Android itself was largely absent from this keynote. Google had already covered Android 17 features, Gemini Intelligence, and a new Googlebook laptop category in a separate Android Show the previous week, freeing the I/O keynote to focus on Gemini and developer tooling; multiple outlets covering the keynote noted Googlebook was not mentioned again on the I/O stage. Developers evaluating the announcements should treat Gemini 3.5 Flash and Antigravity 2.0, both available immediately, as the concrete near-term integration targets, while Spark, Android Halo, and the Android XR audio glasses (launching this fall with Samsung, Warby Parker, and Gentle Monster) remain previews without public API or pricing details yet.

The audio glasses, a lighter-weight companion to Google's previously announced display glasses, ship this fall; watch for developer SDKs or partner announcements as that date approaches. On the agent-payments side, AP2 adoption by merchants and card networks will determine whether Universal Cart's agentic checkout becomes broadly usable. And with three separate agent products (Spark, Antigravity, Search's information agents) now shipping from one company, expect scrutiny over how Google differentiates and eventually consolidates them.