In-Depth Coverage of Google I/O: Gemini 3.5 Takes the Lead, AI Upgrades Across All Product Lines, and Smart Body Ecosystem Rollout

At the 2026 Google I/O conference, Google CEO Sundar Pichai stated that Gemini 3.5 and Antigravity are ushering in a new world of agents and agent capabilities.

Gemini Model Series: Dual Model Division of Labor, Adapting to Large-Scale Implementation and Advanced Generation

Google has released the first member of the Gemini 3.5 model family, Gemini 3.5 Flash.

This main model promoted by Google this time is focused onlow latency, high speed, low cost, and strong agent execution capabilities. The model's comprehensive performance has been greatly improved, achieving breakthroughs in code generation, long-term complex task processing, and multimodal understanding dimensions, setting new highs in various benchmark tests, with output speed four times faster than industry-leading models.

Positioned for large-scale commercial implementation, this model has become the default model for Google's full range of C-side, developer, and enterprise products, covering three major scenarios: Gemini App, Google Search AI mode, and other user products; Google Antigravity, AI Studio, Android Studio, and other developer tools; Gemini enterprise agent platform and other B-side services.

Officially disclosed, the high-end versionGemini 3.5 Pro is currently in internal testing and will be officially released to the public in June 2026, taking on high-end complex AI task demands.

Google's flagship multimodal generation model, Gemini Omni Flash, has made a core breakthrough infull link multimodal input and output. It breaks the traditional limitations of AI only supporting multimodal input and text output, and can generate high-quality video content based on arbitrary material input such as text, images, audio, and video. It will gradually open image and audio output capabilities.

The core functions include: natural language interactive video editing, supporting multiple rounds of modifications while maintaining character, scene, and physical logic continuity; combining world knowledge and physical rules to generate logically rigorous narrative videos; supporting custom voices, and digital avatar creation. All generated content is equipped with SynthID seamless digital watermarking, which is traceable and verifiable.

The launch schedule starts this week with free access to YouTube Shorts, YouTube Create; fully open to AI Plus, Pro, Ultra subscribed users; and in the coming weeks, it will be opened to developers and enterprise customers through APIs.

Gemini App: New 24/7 Personal Agents, Rapid Growth in User Scale

Data shows that the monthly active users of Gemini App have doubled within a year, surpassing900 million, covering 230 countries and regions, and over 70 languages. This update focuses on product experience upgrades and the implementation of agent capabilities.

On the product level, the app features a brand-new Neural Expressive design language, optimizing animation, visuals, and interactive feedback, integrating Gemini Live to achieve seamless switching between text and voice dialogue, and supporting diversified answer outputs such as adaptive text and graphics, dynamic charts, and narrative videos.

Functionally, two core agents are launched: one isDaily Brief, a personalized morning briefing agent that automatically integrates and organizes users' core information every day; the other isGemini Spark 24-hour personal agents that can manage schedules, files, emails, and other full-dimensional digital lives under user authorization.

The launch schedule: Gemini Spark first starts internal testing, then opens the Beta version to AI Ultra users in the United States next week; subsequently, it will be launched on the macOS version of Gemini App, supporting local device task execution. At the same time, Google upgrades the AI subscription system, introducing a 100 USD/month developer subscription plan, and reduces the AI Ultra monthly fee from 250 USD to 200 USD.

Google Search: The Largest Overhaul in 25 Years, Upgrading from Information Search to Task Agent

Google has completed a structural upgrade of its core search business, completely rewriting the traditional search form, and officially entering theSearch Agent era. The site-wide AI mode is equipped with Gemini 3.5 Flash model by default.

On the interaction level, the search box supports dynamic expansion, adapting to complex natural language questions, providing AI intelligent recommendation questions, opening text, images, files, video, browser tabs multimodal input, and supporting continuous dialogue and follow-up