OpenAI has released a research preview of its new Operator AI agent, which is expected to “go to the web to perform tasks for you.” According to a blog post, the tech firm says the agent will be able to interact with its own browser by “typing, clicking, and scrolling.”
The feature is available to ChatGPT Pro tier users in the US, which costs $200 per month. That said, as it is still in research mode, OpenAI acknowledges that there will be limitations and that it will continue to evolve based on user feedback. The company says it’s planning to bring the tool to more users in its Plus, Team, and Enterprise tiers down the line.
A research preview of Operator, an agent that can use its own browser to perform tasks for you. pic.twitter.com/wkBBDIlVqj
— OpenAI (@OpenAI) January 23, 2025
During a livestream, CEO Sam Altman stated that “[Operator] will be [in] other countries soon,” but admitted that it would “take a while” for it to be rolled out in Europe.
The initial research preview is available at operator.chatgpt.com, but OpenAI says it’s planning to integrate Operator into all of its ChatGPT apps soon.
OpenAI says it is collaborating with companies like DoorDash, Instacart, OpenTable, Priceline, StubHub, Thumbtack, Uber, and others to ensure the system “addresses real-world needs while respecting established norms.”
How does OpenAI’s Operator work?
OpenAI explains that Operator runs on a new model called the Computer-Using Agent (CUA), which combines GPT-4o’s vision capabilities with advanced reasoning powered by reinforcement learning. It’s trained to interact with graphical user interfaces (GUIs)—the buttons, menus, and text fields you see on a screen.
With Operator, the model can “see” by analyzing screenshots and “interact” using mouse and keyboard actions. This helps it to navigate the web and take actions without relying on custom API integrations.
If it hits a snag or makes a mistake, it uses its reasoning skills to self-correct. And when it really gets stuck, it hands control back to you.
Featured image: OpenAI