OpenAI Launches Advanced ChatGPT Agent with Autonomous Task Capabilities

OpenAI launches the ChatGPT agent, an advanced AI capable of autonomous task execution.

Key Points

  • • ChatGPT agent can perform tasks autonomously using a virtual computer.
  • • Available to Pro, Plus, and Team users with plans for broader access.
  • • Achieved a benchmark score of 41.6 on Humanity's Last Exam and 27.4 on FrontierMath.
  • • Enhanced safety measures implemented to manage potential risks.

OpenAI has officially launched its new ChatGPT agent, a groundbreaking development in AI technology designed to autonomously perform complex tasks using its own virtual computer. Announced on July 17, 2025, this agent represents a major evolution in the capabilities of AI applications, permitting users to activate 'agent mode' to handle activities such as scheduling meetings, generating presentations, or even managing web interactions.

The ChatGPT agent combines features from previous versions, including the web interaction capabilities of the Operator and the robust information synthesis of Deep Research, allowing for seamless transitions between both reasoning and practical action based on user instructions. Users of the Pro, Plus, and Team plans can begin utilizing this innovative agent, with plans for broader access to be expanded later.

Performance benchmarks reveal that the ChatGPT agent excels with a score of 41.6 on Humanity’s Last Exam, indicating a notable increase in effectiveness compared to earlier models. In addition, it achieved a score of 27.4 on the tougher FrontierMath benchmark. This enhanced performance is reportedly due to its ability to integrate tools like visual and text-based browsers and APIs, which facilitate real-time web interactions and automate tasks including API calls and form submissions

Notably, OpenAI has addressed potential risks associated with the ChatGPT agent. Enhanced safety measures aim to mitigate vulnerabilities, particularly in terms of prompt injection and sensitive data handling, with explicit user confirmations required before critical actions are undertaken. OpenAI has also disabled the memory feature, designed to protect user data from being exfiltrated, ensuring a safer user experience.

As this agent integrates varied capabilities, it is seen not only as a tool for individual users but also targeted towards enterprise customers, streamlining operations that could improve organizational productivity. The company’s focus on refining the ChatGPT agent aligns with industry trends, where major tech firms are investing heavily in AI agents to enhance efficiency.

While the initial rollout is limited to paying subscribers, the ongoing evolution of the ChatGPT agent indicates OpenAI’s commitment to lead in the competitive AI landscape. Continuous improvements are expected as user feedback helps refine its functionalities and enhance user experience in practical applications like managing task automation effectively.