Google I/O 2026 Major Release: Gemini 3.5 Ushers in Agentic AI Era, Search Gets Biggest Upgrade in 25 Years

Gemini 3.5's release was the highlight of this I/O conference. According to Google's official introduction, Gemini 3.5 is a frontier model designed specifically for the agentic AI era, surpassing all competitors in multiple benchmark tests. Compared to the previous Gemini 3, version 3.5 improved code generation capability by 40% and accuracy on complex reasoning tasks by 35%. More importantly, Gemini 3.5 possesses true 'agentic' capabilities — it can not only answer questions but also proactively execute multi-step tasks, call external tools, and adjust strategies based on intermediate results. For example, users can tell Gemini 'help me plan a Japan trip and book hotels,' and Gemini will automatically search for flights, compare hotel prices, check visa requirements, and even complete bookings directly. This shift from 'passive answering' to 'proactive execution' marks AI assistants entering an entirely new capability tier.

Google Search's upgrade is equally remarkable. Pichai called it 'the biggest transformation in search in 25 years.' The traditional search model involves users entering keywords and search engines returning link lists. The new search model introduces the 'information agents' concept — these AI agents work continuously in the background, proactively monitoring information domains users are interested in. For example, users can set 'monitor Tesla stock price and related news,' and information agents will track relevant information 24/7, proactively pushing detailed updates and in-depth analysis when important changes are discovered. Another revolutionary feature is Antigravity — it can transform search results into dynamic, interactive visual interfaces. When users ask complex questions, Antigravity doesn't return text links but generates customized dashboards, interactive charts, or even mini applications. This shift from 'searching for information' to 'generating experiences' completely redefines the search engine's role.

On the hardware front, Google also brought surprises. At the Android Show, Google announced the Googlebook — a new device jointly launched with hardware partners, optimized specifically for AI agents. Although specific specifications haven't been fully disclosed, Google emphasized that Googlebook is specially optimized for Gemini 3.5's agentic functions, capable of running complex AI tasks more efficiently. Additionally, Fitbit Air's release marks Google's major layout in personal health. Fitbit Air not only has traditional wearable device health monitoring functions but also integrates an AI health coach that can provide personalized health advice and exercise guidance based on users' physiological data. Combined with the newly launched Google Health app, users can enjoy comprehensive AI health services ranging from symptom analysis and medical record interpretation to personalized health guidance. Google states these health tools aim to 'give everyone a private AI health advisor.'

From a strategic perspective, Google's release demonstrates its comprehensive layout in the AI race. Gemini 3.5 represents breakthroughs in foundational model capabilities, the search upgrade showcases AI application innovation in core products, and health tools reflect AI's deep penetration into vertical domains. Notably, Google repeatedly emphasized 'responsible AI' and 'safety first' in its release — contrasting sharply with the consecutive security incidents at OpenAI and Anthropic recently. Google stated all new features underwent rigorous safety testing and established multi-layer protection mechanisms. Analysts believe Google's release is not only a response to competitors but also a definition of AI's future development direction. Through the 'agentic AI' concept, Google attempts to elevate AI from 'tool' to 'partner,' from 'passive response' to 'proactive service.' If this strategy succeeds, it will fundamentally transform how humans interact with technology.