On May 19–20, 2026, Google’s annual I/O developer conference concluded in Mountain View, California. This I/O, themed "Welcome to the Agentic Gemini Era", marked a comprehensive shift toward Agentic AI. Google CEO Sundar Pichai and DeepMind CEO Demis Hassabis both emphasized that AI has evolved from an "assistant" into agents capable of autonomously planning and executing complex tasks. The keynote unveiled major updates across models, developer platforms, Search, Android and hardware, showcasing Google’s ambitions in the multimodal and agent era.

Key highlights: Gemini 3.5 series and Gemini Omni

Gemini 3.5 Flash was the first major model announced. While maintaining state-of-the-art capabilities, it delivers a significant speed boost and is described by Google as "the most capable agentic and coding model to date." Compared with its predecessor, 3.5 Flash shows markedly improved performance on long-horizon, multi-step reasoning tasks, produces outputs at roughly four times the speed of other leading models, and consumes fewer tokens, offering a substantially better cost-performance ratio. The model is now available globally and is the default for Gemini App and Search AI Mode.

Gemini Omni was the most striking new multimodal entrant at the event. Positioned as a "world model", Omni can take arbitrary inputs (text, images, video, audio) and produce arbitrary outputs, with an initial focus on video generation and conversational editing. Users can iteratively modify video content through natural-language instructions — for example, "make the character’s movements more natural," "replace the background with a sunset," or "add a voice-over narration" — greatly lowering the barrier to video creation. The Omni Flash tier has been opened to subscribers, with image and audio capabilities to follow.

Google also introduced Antigravity 2.0, a platform dedicated to agent development that helps developers rapidly build, orchestrate and manage agents while integrating security mechanisms such as endpoint sandboxes and credential protection.

Android 17 and deeper ecosystem integration

At I/O, Google detailed the AI features in Android 17, centered on Gemini Intelligence and its multi-step automation abilities. Users will be able to use natural language to have Gemini perform complex cross-app tasks, such as extracting information from emails, placing orders automatically, and creating calendar events. These capabilities will roll out first to Pixel 11 and the next-generation Samsung Galaxy devices.

Google also showcased the Samsung-collaborated Android XR smart glasses that integrate Gemini capabilities, supporting real-time translation, visual search and AI-assisted navigation, further extending AI use cases on wearable devices.

Major upgrades to Search and Workspace

Search received a "big update." AI Mode now covers more than one billion monthly active users, and Google demonstrated generative UIs and agent-style search features that let users complete more complex queries and tasks through conversation. Workspace also gained multiple new AI features to boost productivity.

Key takeaways from the keynote

Agentic AI becomes the central theme — Google has framed 2026 as the "Agentic Gemini Era." From Gemini 3.5 Flash to Antigravity 2.0 and agent capabilities in Search and Android, Google is pushing AI from passive response toward proactive execution.
Breakthroughs in multimodality and world models — The launch of Gemini Omni signals meaningful progress in multimodal understanding; future AI systems will have stronger real-world perception.
Developer tools upgraded across the board — Tools such as Google AI Studio, Antigravity