On May 20, the Google I/O 2026 Developer Conference arrived as scheduled.
1. Gemini 3.5 Flash: Faster, Cheaper, Available to Everyone
The core release of this I/O is undoubtedly Gemini 3.5 Flash.
In terms of performance, Pichai gave a set of intuitive data: the output speed of Gemini 3.5 Flash is faster than other cutting-edge models. 4 times, its performance comprehensively surpasses Gemini 3.1 Pro in coding and agent tasks, while the price is less than half of the same model.
What is more worthy of attention is thatGemini 3.5 Pro will launch next month, already in use internally at Google.
2. Google search is fully AI-powered: the biggest change in 25 years
If the model upgrade is expected, then the search revision really makes people feel that "the times have changed."Google search box will be fully powered by Gemini 3.5 Flash, what users see after entering a query is not a traditional blue link list, but a customized summary page generated by AI.
Specific changes include:
- Conversational search: You can ask questions and discuss in depth just like chatting, and the search experience changes from "one-time query" to "continuous dialogue"
- multimodal input: In addition to text, you can also use images, videos, files, and even Chrome tabs to ask questions
- "Information Agents": Users can set up an AI agent to continuously monitor specific topics in the background. For example, after setting the conditions for finding a house, the agent will automatically scan the property and notify you when it matches.
Currently, AI Overviews monthly active users have exceeded 2.5 billion, Elizabeth Reid, vice president of search at Google, said that the new search allows "context to flow with you, and the deeper you go, the more precise it becomes."
3. Gemini Spark: Your 24-hour personal AI assistant
The most imaginative product at this conference is Gemini Spark—A persistent AI agent running on a Google Cloud virtual machine.
Pichai’s original words were: “You don’t have to have your laptop open all the time, it’s online 24 hours a day.” The first batch will be open to trusted testers, and will later be rolled out to Google AI Ultra subscribers in the United States. Android Halo With this new interface, users can also view the running status of the background agent in real time on their mobile phones.
4. Gemini Omni: a new paradigm for AI video generation
In terms of creative tools, Google released Gemini Omni Flash——A "world model" level video generation tool.
The best thing is,Omni Flash is free for all users starting this week, integrated directly into YouTube Shorts and the YouTube Create App.
5. Security and Ecology: SynthID exceeds 100 billion, OpenAI also joins
On the AI safety front, Google announced its SynthID watermarking system has tagged over 100 billion AI-generated images and videos, and tens of thousands of years of audio content.OpenAI, Kakao and ElevenLabs Officially join the SynthID ecosystem, which means that cross-platform AI content tracing will become a reality in the future.
In addition, Google revealed that capital expenditures in 2026 will reach US$180-190 billion(only 31 billion in 2022), and released the eighth-generation TPU chip. 900 million, 8.5 million developers call the Gemini API every month, processing approximately 19 billion tokens every minute.
Summary: AI changes from “tool” to “agent”
Throughout the entire Google I/O, the core keyword is "Agentic"——AI is no longer a tool that passively waits for instructions, but an intelligent agent that actively understands tasks, continues to execute them, and acts on behalf of the user.
For ordinary users, Gemini 3.5 Flash and Omni Flash can already be experienced for free; for developers and enterprises, lower prices and higher efficiency mean that the threshold for AI implementation is rapidly lowering.
Want to discover more useful AI tools? AI Dash ——Discover the best AI tools.