Google unveils Gemini Omni and Gemini 3.5 at I/O 2026

Google unveiled two new artificial intelligence models at I/O 2026 and announced an expansion of agents capable of carrying out tasks, not just answering questions. The company defines this stage as the era of Gemini agentic AI.

The main new releases are Gemini Omni and Gemini 3.5 Flash. According to Google, Omni can create content from any type of input, starting with video, while improving its understanding of the world, its ability to work across formats and its editing capabilities.

Gemini 3.5 Flash belongs to Google's new model family and combines advanced reasoning with the ability to take action. In practice, the difference is that the model does not stop at giving you an answer: it can also help you complete steps within a task.

From answering to acting

Google also showed progress on Antigravity, its platform for developing applications with agents from the start. An agent is an AI system that can break a goal into steps, use tools and carry out actions with less human intervention.

This changes how you use these tools. Instead of asking AI to draft an email, for example, the goal is for it to help you prepare the reply, find the necessary information and move the task forward. Google says this approach will allow more people to create applications, even without traditional programming experience.

Agents inside Google's products

The company is also bringing these capabilities to its services. The announced experiences include:

• Information agents in Google Search, to research and organize answers with greater autonomy.
• Gemini Spark and Daily Brief within the Gemini app.
• Universal Cart, a shopping cart designed to compare and manage products more intelligently.
• Ask YouTube, to interact with the platform's content through questions.

Google says it is also bringing Gemini to more products and devices, including Google Pics, smart glasses and other hardware formats. The goal is for AI not to live only in a chat application, but to appear wherever you carry out a task.

What changes for you

For now, the announcement describes a product direction and new Google features, not universal and immediate access to all of them. The important change is the shift in focus: the company wants Gemini to move from being an assistant you consult to becoming a system that can help you complete processes.

That will make availability, permissions and oversight just as important as the model's capabilities. The next question will not only be what Gemini can generate, but what actions it can take, with which data and under what user controls.