ChatGPTcreator OpenAI has finally entered the agentic AI slipstream with the passing of its Operator AI in January . The agentic arrangement is designed to lick autonomously on its user ’s behalf and is primed to compete against already established industry rivals likeClaude ’s Computer Use APIandMicrosoft ’s Copilot agents — at least , once it sheds its “ research trailer ” status .
Here ’s everything you need to know about OpenAI ’s young agent and when you might be able to adjudicate it for yourself .
What is Operator?
OpenAI ’s Operator isan agent AI , meaning that it is plan to take self-governing action based on the information uncommitted to it . But unlike conventional programs , AI agents are able to review changing conditions in veridical - time and react consequently , rather than merely fulfil preset commands . As such , AI agents are capable to perform a variety of complex , multi - step tasks run from transcribe , summarizing and give action items from a concern coming together to book the trajectory , hotel accommodation , and rental car for an forthcoming holiday based on your household ’s various schedules to autonomously search topics and assembling multi - page report on those subjects .
wheeler dealer works slimly differently than other factor presently available . While Claude ’s Computer Use is an API and Microsoft ’s AI agents work within theCopilotchat UI itself , Operator is designed to , well , operate , within a consecrated web web web browser windowpane that runs on OpenAI ’s servers and executes its chore remotely . Your local web web browser has nothing to do with the process and can be used commonly even when Operator is running .
The Operator app is power by a new “ reckoner - Using Agent ” model ( CUA ) that is , in turn of events , built atopGPT-4o , which provides the app ’s multimodal abilities . OpenAI tell CUA was rail in a exchangeable fashion to its o1 and o3 abstract thought models . As such the CUA model will break down complex task into their component problems before trying to lick them sequentially , backtrack if it move into any logical roadblocks .
When was Operator released?
OpenAI free Operator on January 23 , 2025 . It ’s presently only usable to $ 200 / month Pro users in the U.S. “ Our plan is to expand to Plus , Team , and endeavour drug user and integrate these capability into ChatGPT in the future , ” the companywrote in its announcement post .
How does Operator work?
Users can activate the agent from the ChatGPT place screen , which pop up a dedicated web internet browser page in a side window for Operator to contain out its tasks . The AI provides a track story of what it is currently doing and the exploiter can take over the cognitive operation at any prison term . Operator will call for for the substance abuser ’s assistance in certain job , such as logging in to specific secured websites , as well as get the substance abuser ’s verification before executing important tasks . It can interact with site both visually ( i.e. through screen shots ) and tactically , when it mimics the user ’s keyboard taps and computer mouse click .
What can Operator do and how well can it do it?
Since it is specify to the web web browser , Operator can only perform simple internet - based project currently , such as appropriate concert tag , ordering DoorDash , or filling out Instacart orders . The party also arrogate that the factor will be able to automate tasks like book hotel and airlines , reserving tables at eatery , and even doing your online shopping .
OpenAI has play off Operator against Anthropic ’s Computer Use , as well as Google DeepMind ’s Mariner agent , in a identification number of diligence benchmark and claims that Operator has beat them across the control panel . On theOSWorld benchmark , which measure how well an agent can complete tasks like coalesce PDF file cabinet , CUA ticktack out Computer Use 38.1 % to 22.0 % — for reference , humans average around 72 % achiever on those tasks . On theWebVoyager benchmarkCUA outscored Mariner 87 % to 83.5 % . information processing system Use hit a paltry 56 % .
However , initial user reactions to the AI agent have been mixed . For example , New York Timescolumnist Kevin Roost write , “ In all , I find that using Operator was normally more trouble than it was worth . Most of what it did for me I could have done quicker myself , with few vexation . ”
“ Even when it worked , ” he uphold , “ it asked for so many substantiation and reassurances before acting that I felt less like I had a virtual assistant and more like I was supervising the world ’s most insecure medical intern . ”
