OpenAI has just officially launched ChatGPT Agent, a powerful AI actor capable of automating a series of complex tasks for users. This is considered a big step forward in the race for the development of representative AI among technology giants.
Recently, CEO Sam altman and the OpenAI engineering team demonstrated the outstanding capabilities of ChatGPT Agent: from meeting scanning, web browsing, shopping, planning to creating presentation slides.
This tool acts as a powerful virtual assistant, can operate tasks on a virtual machine and automatically select the optimal tool for each task.
ChatGPT Agent is powered by a new, professionally trained AI model to handle complex tasks that require multiple tools such as image browsers, text, API access, and terminal devices.
In particular, it can also connect to applications like Gmail or GitHub through integrated connectors.
OpenAI said that users of the Pro, Plus, and Team packages can experience the Agent feature today by enabling rocher mode in the ChatGPT interface.
At the core of this new system is an architecture called a one-stop worker, combining the capabilities of the two previous Operator and Deep Research tools.
In the performance tests, the ChatGPT Agent support model reached 41.6% in the The Last Test of Humanity (HLE), a challenging academic review.
Other scores included 68.9% in Browse Comp (web browsing capability), 45.5% in spreadsheetBench, and 27.4% in Frontier Math. In particular, in data science tests (DSBench), ChatGPT Agent is superior to humans.
OpenAIs launch of ChatGPT Agent shows the company is accelerating strongly in the AI race, where Amazon, Google and Meta are also investing heavily.
If 2023 is the time when the concept of AI actor becomes popular, 2025, witnessing fierce competition to create a multi-functional AI tool that can truly accompany and support users like a real virtual partner.
