At its annual I/O developer conference on May 19–20 in Mountain View, Google outlined a broad push to embed Gemini agents across its most important surfaces — Search, Android, and a new pair of audio glasses. The message was not about a single product but about stitching those agents into a front-end layer that sits between the user and Google’s entire ecosystem.
The move connects the company’s model releases, its operating system, and its eyewear hardware into one integrated AI platform, turning the developer gathering into a showcase for agentic computing. For the 900 million people already using Gemini each month — a figure cited by Chief Executive Sundar Pichai — the shift signals that the chatbot is becoming a persistent, proactive assistant that reaches far beyond a text box.
Google DeepMind published model cards for two new foundation models: Gemini 3.5 Flash and Gemini Omni Flash. The company claimed the Flash variant was 4x faster than other frontier models in output tokens per second, according to reports from 9to5Google, though benchmark details were not provided. Google also previewed Gemini 3.5 Pro, a reasoning model expected next month, and a lightweight agent called Gemini Spark, according to multiple outlets covering the keynote. Gemini Intelligence for Android will bring multi-step task automation to recent Samsung Galaxy and Google Pixel phones starting this summer, extending later to watches, cars, glasses and laptops. This agent can operate across apps without constant user prompting.
Google showed a new intelligent search box that can surface dynamic, generative UI — layouts customized to the query — and background information agents that can carry out research tasks without a user present. The overhaul will roll out first to subscribers of Google AI Pro and Ultra tiers, with broader access to follow. Specific launch dates were not disclosed, and independent reports suggest the timeline remains staggered. This marks a deeper step toward a Search experience that answers questions with structured, AI-generated layouts rather than a list of links.
Google officially announced its first Android XR audio glasses, developed with Samsung, Warby Parker and Gentle Monster. The frames are scheduled to launch this fall, with display-enabled glasses to follow later. No price was given, and the audio glasses will pair with both Android and iOS. This is not Google’s first foray into smart eyewear — it previewed Android XR glasses at I/O 2025 — but the 2026 version brings specific release timing and retail partners, moving from concept to near-term product.
Pichai told attendees that the company now has 2.5 billion monthly users of AI Overviews, about 1 billion of AI Mode, and more than 50 billion images generated with Gemini, according to Reuters and Wired. Those numbers are not yet confirmed by a company filing or transcript. Separately, reports from Reuters, TechCrunch and Axios describe a restructuring of the Google AI Ultra subscription: a $100 plan and a $200 plan replacing a prior $250 price point. The company did not immediately confirm the new pricing structure.
For developers, the announcements offered a tighter link between Gemini models, Android XR, and Search — but availability and pricing details for the APIs remain unclear. Google is competing with OpenAI, Anthropic, and Meta, which have all pushed agentic features and wearable AI devices. By owning the OS, the search front-end, and the hardware reference designs, Google is positioning its stack as the lowest-friction path for agents to reach consumers. However, much of the showcase consisted of future promises, with actual delivery staggered across months and subscriber tiers.
Google’s platform play at I/O 2026 was unambiguous: make Gemini the connective tissue across devices and interfaces. Execution and rollout speed will determine whether that integrated ecosystem materializes before developers and users shift their attention elsewhere.