After launching Deep Research and Opeartor agents earlier in the year, OpenAI has launched a new AI agent called ‘ChatGPT agent’ that more deeply intergrates by the company’s chatbot with the same name. ChatGPT agent is general purpose agentic system that can take actions on the user’s behalf.
With the new ChatGPT agent, the chatbot will be able to handle tasks like accessing the user’s calendar, shopping on their behalf, create spreadsheets and even browse the web to gather information. Furthermore, it can also be used to automate repetitive tasks like converiting screenshots or dashboards into presentaions, planning and booking offsites, updating spreadshets with new data and more.
OpenAI says ChatGPT agent brings the best of both world from Operator and Agent and it can naturally transition from a simple conversation to requesting actions directly within the same chat.
ChatGPT can access your Gmail and Github:
While there were rumors about OpenAI launching its own AI browser after the release of Perplexity’s Comet, the company has provided a range of new browsing capabilities in the ChatGPT agent which it says is a “visual browser that interacts with the web through a graphical-user interface, a text-based browser for simpler reasoning-based web queries, a terminal, and direct API access.”
ChatGPT agent can also take advantage of ChatGPT connectors to access websites like Gmail and Github in order to find relevant information around the prompt and use it in its response. Users wil also have the option to log into different websites at any point by taking over the browser.
The new agent manages different tasks by leveraging its own virtual computer which OpenAI says “preserves the context necessary for the task, even when multiple tools are used”.
The ChatGPT agent is said to be more flexible and interactive than previous models and userscan interrupt it any point to clarify their intructions, or steer it towards a desired outcome and even change the task entirely. Moreover, the agent may seek additonal details from the user that may be needed in order to complete the task.
ChatGPT agent is better at managing spreadsheets than Copilot:
In terms of benchmarks, OpenAI says that ChatGPT agent scores higher than o3 reasoning model and Deep Research agent on Humanity’s Last Exam. Whereas on FronteirMath, a math benchmark featuring novel, unpublished problems, it has better accuracy than o3 and o4 mini.
Interestingly, on SpreadsheetBench, a benchmark which evaluates models based on their ability to edit spreadsheets derived from real-world data, OpenAI says ChatGPT agent not only outperforms its own models also Microsoft’s Copilot.
OpenAI CEO Sam Altman called the launch of ChatGPT agent a “feel the agi” moment in a post on X. He noted, “something about seeing the computer think, plan, and execute hits different.”
How to use ChatGPT agent?
ChatGPT agent can be accessed by selecting ‘agent mode’ in the ChatGPT app or website. The feature, however, is only available to OpenAI’s paying members which includes Pro, Plus and Team users at this point.
