Google I/O ’26 Keynote

The theme of Google I/O 2026 was unmistakable: Google officially declared the transition from conversational chatbots to the “Agentic Gemini Era.” The two-hour keynote at the Shoreline Amphitheatre focused entirely on AI that doesn’t just answer questions, but proactively acts, operates in the background, and seamlessly cross-references data across tasks.

Here is the comprehensive breakdown of the major announcements from the I/O ’26 keynote.

1. Next-Gen Frontier Models: Gemini 3.5 & Omni

Google introduced two completely new model families, shifting its focus toward raw execution speed and multi-modal creation.

• Gemini 3.5 Flash: Rolling out immediately as the default model powering the Gemini app and Google Search. It clocks in at 4x faster output tokens per second than rival frontier models and actively outperforms Gemini 3.1 Pro on coding, tool-handling, and complex visual benchmarks. (A heavier Gemini 3.5 Pro variant is scheduled for next month).

• Gemini Omni: A highly anticipated multimodal “world model” that blends text, images, video references, and audio into realistic media. Launching initially as Gemini Omni Flash, it lets paid subscribers generate and conversationally edit 10-second cinematic video clips directly inside a new “Videos” tab in the Gemini app and Google Flow, with a free rollout hitting YouTube Shorts.

2. The Search Reimagining: Generative UI & Search Agents

Google Search received its most radical overhaul since its inception, moving away from simple link aggregation into a dynamic, interactive workspace.

• The Intelligent Search Box: The standard search box now dynamically expands as you type. Instead of classic autocomplete, it uses AI to help you articulate hyper-specific requests. You no longer have to manually toggle “AI Mode”—the system seamlessly integrates tools based on your intent.

• Search Agents: Users can now deploy 24/7 “Information Agents” right inside Search. For instance, you can task an agent with tracking a complex topic or upcoming update; it will continuously monitor the web in the background and ping you with organized, synthesized digests.

• Direct Execution: You can now drop code directly into the search bar for debugging or ask Search to visually map out concepts using live, interactive on-screen graphics.

3. Meet Your 24/7 Workspace Partner: Gemini Spark

Taking the concept of agents to the desktop, Google introduced Gemini Spark, an autonomous personal assistant built to handle long-horizon workflows across the Google Workspace ecosystem.

• Background Automation: Spark runs on dedicated virtual machines in Google Cloud, meaning it operates continuously 24/7 even if your laptop is closed or your phone is locked.

• Cross-App Orchestration: Rather than waiting for prompt-by-prompt instructions, Spark can ingest a broad objective (e.g., “Prepare for tomorrow’s client meeting”) and execute a chain of events: pulling histories from CRM tools, drafting an account strategy document in Docs, updating timelines in Sheets, and queueing up a confirmation email in Gmail awaiting your final approval.

4. Hardware & Developer Ecosystem: TPU v8 & Antigravity 2.0

To power this massive agentic shift, Google showcased its upgraded backend infrastructure and new developer toolkits.

• TPU v8 (Dual-Chip Architecture): For the first time, Google split its Tensor Processing Units into specialized hardware: TPU 8t (optimized for massive, distributed global model training across over 1 million chips) and TPU 8i (engineered for ultra-low latency inference to handle continuous agentic loops).

• Google Antigravity 2.0: A standalone desktop app and CLI built for “agent-first” development. It allows engineers to orchestrate specialized subagents to concurrently code, test, and debug pipelines. It features advanced sandboxing, credential masking, and seamless exporting from Google AI Studio.

• Android Studio Migration Agent: A new AI utility capable of automatically converting legacy mobile codebases or cross-platform web frameworks directly into native Kotlin Android applications.

5. Wearables & New Creative Hardware

• Android XR No-Screen Audio Glasses: Developed in partnership with Samsung and Qualcomm—with frames designed by Warby Parker and Gentle Monster—these sleek, screenless smart glasses launch this fall. Users tap the frame or say “Hey Google” to navigate, capture media, run real-time audio translations that match the speaker’s vocal pitch, and receive contextual audio cues about their surroundings.

• Google Pics: A new standalone app and Workspace integration dedicated exclusively to advanced, agentic image generation and design layer management inside Google Slides and Drive.

• SynthID Expansion: Google announced deeper integration of SynthID and C2PA Content Credentials directly into Google Search and Chrome, allowing everyday users to instantly verify whether an image or video’s origin is authentic camera capture or AI-generated/edited.