Everything Announced at Google I/O 2026

At the center of Google I/O 2026 was less about model announcements and more about development flows where agents step directly into products, browsers, cloud, mobile apps, data, and testing environments.

Key model and platform updates are as follows.

• Gemini 3.5 Flash: a new core model that puts fast execution and cost efficiency front and center
• Gemini Omni Flash: the first model in the Omni line, handling a broader range of multimodal input and output
• Gemma 4: an open model family available in 2B–31B sizes, under the Apache 2 license
• Antigravity 2.0: a standalone desktop app that bundles agent conversations, projects, and task management
• Google AI Studio: expanded from idea generation all the way to Cloud Run deployment, Workspace integration, and Android app creation

Google emphasized a direction where developers don't stop at writing code, but hand off planning, execution, verification, debugging, deployment, evaluation, and multi-agent orchestration as well. The Antigravity demo showcased generating the core of an operating system at a scale of 93 sub-agents, over 15,000 model requests, and 2.6 billion tokens.

Infrastructure and developer tools are also shifting to be agent-centric.

• Managed agents are being added to the Gemini API, letting harnesses like Antigravity be used in API form
• Access to Google Cloud tools and data through 50+ managed MCP servers
• Developer Knowledge MCP supplies snapshots of the latest documentation at roughly 8–12 hour intervals
• Data Agent Kit and BigQuery MCP enable Firestore/BigQuery/log analysis and dashboard generation
• A remediation agent and CI agent tying together Cloud Run, Firebase, Eventarc, and Pub/Sub handle error investigation and automated recovery

Android, Chrome, Firebase, and Play are aligned in the same direction.

• Android 17: stronger checks on quality issues such as excessive memory/CPU usage and cold start
• Strengthened focus on Jetpack Compose, with Compose 1.10/1.11 and expanded Navigation 3 for large-screen devices
• Chrome: Prompt API and Modern Web Guidance provided so agents can correctly use the latest Baseline-based web features
• Firebase: transitioning into an agent-native platform where humans and agents build apps together
• Google Play: expanded payment methods, a subscription management API, and enhanced app discovery via Gemini

Overall, rather than competing on model performance, Google put forward a system for safely connecting the latest models into actual products and development flows as the bigger theme.