Google’s Gemini 2.5 Computer Use: An AI That Can Operate Your Web Browser

Google has launched a groundbreaking new AI model named Gemini 2.5 Computer Use, currently available in preview for developers. This specialized model represents a significant leap in AI capabilities, enabling artificial intelligence to see, understand, and interact with web interfaces just like a human user.

What Is Gemini 2.5 Computer Use?

Built upon the powerful Gemini 2.5 Pro foundation, Computer Use is a specialized model designed to power AI agents that can interact with graphical user interfaces (UIs). Its core function is to perform tasks in environments where a direct API connection isn’t available, requiring it to navigate the visual elements of a website or application.

This means the AI can be instructed to:

Scroll through web pages
Click on buttons and links
Fill out and submit forms
Drag and drop items
Navigate menus

How It Works: The AI Interaction Loop

The model operates through a continuous, intelligent loop:

Analyze: It looks at the user’s prompt, takes a screenshot of the current screen, and reviews a history of recent actions.
Act: It performs the next logical step in the task (e.g., clicking a button).
Repeat: It captures a new screenshot to see the result of its action and then continues the process until the task is complete.

A demo video showcases the AI organizing a chaotic digital sticky note board by dragging notes into the correct categories, demonstrating a sophisticated understanding of visual context and instructions.

Current Availability and Performance

For Developers: The model is currently available for testing via the Gemini API in Google AI Studio and Vertex AI, as well as through the platform Browserbase.
Not for General Users: It is not yet available to the public within the standard Gemini chatbot.
Top-Tier Performance: Google’s benchmarks indicate that Gemini 2.5 Computer Use demonstrates “strong performance” on multiple web and mobile control benchmarks, potentially outperforming rival models like OpenAI’s Agent.

Important Note: Unlike some competitors, Google’s model is currently optimized specifically for browser control and is not designed for general operating system navigation.

The Future of AI Agents

The release of Gemini 2.5 Computer Use is a foundational step toward developing sophisticated AI agents. This technology could eventually be integrated into future versions of the Gemini assistant in Chrome and Google Search, empowering them to complete complex, multi-step tasks on behalf of users directly within their web browser.