OpenAI Launches Operator, a Browser-Based AI Agent for Task Automation

January 23, 2025

OpenAI has introduced Operator, a groundbreaking browser-based AI agent designed to perform tasks on the web. This research preview is now available to Pro users in the U.S., accessible at operator.chatgpt.com. Operator can handle a variety of repetitive and time-consuming tasks, from filling out forms to ordering groceries, by interacting with web pages just like a human user.

What Changed Technically?

Operator is powered by a new model called the Computer-Using Agent (CUA). This model combines GPT-4's advanced vision capabilities with sophisticated reasoning through reinforcement learning. Here’s a breakdown of how CUA and Operator work together:

• Vision Capabilities: CUA can "see" web pages through screenshots, allowing it to understand the layout and content of any webpage.
• Interaction: It can perform all actions typically done by a mouse and keyboard, such as clicking, typing, and scrolling. This means Operator doesn’t need custom API integrations to interact with websites.
• Reasoning and Self-Correction: If Operator encounters challenges or makes mistakes, it uses its reasoning capabilities to self-correct. In cases where it gets stuck, it can hand control back to the user for assistance.

Why It Matters to Practitioners

Operator represents a significant step forward in AI's ability to interact with web interfaces. For developers and businesses, this means:

• Time Savings: Automating repetitive tasks can free up valuable time for more complex work.
• Broader Utility: By using the same tools and interfaces humans use daily, Operator broadens the scope of what AI can do, making it more accessible and practical.
• Iterative Improvement: The research preview nature of Operator allows OpenAI to gather user feedback and continuously refine the model.

Key Features and Capabilities

• Task Execution: Operator can handle a wide range of tasks, including:
  • Filling out forms
  • Ordering groceries
  • Creating memes
• User Collaboration: When it encounters difficulties, Operator hands control back to the user, ensuring a smooth and collaborative experience.
• State-of-the-Art Performance: CUA sets new benchmarks in WebArena and WebVoyager, two key browser interaction datasets.

Availability and Future Plans

Operator is currently available as a research preview to Pro users in the U.S. OpenAI plans to expand access to Plus, Team, and Enterprise users in the future. Additionally, Operator will be fully integrated into ChatGPT as "agent mode" by selecting it from the dropdown menu in the composer.

Recent Update

As of July 17, 2025, Operator is now fully integrated into ChatGPT. To access these updated capabilities, users can select “agent mode” from the dropdown in the composer and enter their query directly within ChatGPT. The standalone Operator site (operator.chatgpt.com) will be sunsetted in the coming weeks.

Conclusion

The launch of Operator marks a significant advancement in AI's ability to interact with web interfaces. By combining advanced vision capabilities with sophisticated reasoning, OpenAI has created a powerful tool that can help users save time on everyday tasks and open up new opportunities for businesses. As the model continues to evolve based on user feedback, we can expect even more robust and versatile capabilities in the future.