OpenAI Unveils Unified ChatGPT Agent with Enhanced Capabilities

OpenAI has announced the release of its unified ChatGPT Agent, which combines the capabilities of its Operator and Deep Research tools into one platform. The new agent is designed to handle complex tasks, make informed decisions about which models to use, and navigate the web independently.

During a live stream event on July 17, 2025, OpenAI CEO Sam Altman revealed that ChatGPT Agent is a unified agentic model that can decide for users which tool would best suit their query. The agent is trained using reinforcement learning and has access to a range of tools, including a terminal to write code and other ChatGPT APIs.

ChatGPT Agent offers several key features, including the ability to interrupt tasks mid-stream, ask for confirmation before completing steps, and retrieve evaluation numbers from Google Drive. However, OpenAI notes that the agent may be vulnerable to malicious websites and emphasizes the importance of using “takeover mode” to mitigate potential risks.

The new ChatGPT Agent is rolling out today for Pro, Team, and Plus users, with 400 queries a month available for Pro users and 40 for Team and Plus users. OpenAI expects the rollout to be completed soon for Team and Plus users.

While the name of the agent remains unclear, with OpenAI using “ChatGPT Agent” with an uppercase “A,” it appears that “agent mode” is the correct terminology. The new capabilities offer a significant upgrade to ChatGPT and expand its functionality in several ways.

As part of its release, ChatGPT Agent aims to provide a highly collaborative experience for users, offering features such as task interruption, confirmation requests, and evaluation number retrieval. With its integration with other tools and models, ChatGPT Agent is poised to become a powerful tool for individuals looking to leverage AI capabilities in their daily lives.

Source: https://www.techradar.com/news/live/openai-july-17-announcement-live-event