OpenAI introduces “ChatGPT Work,” a new agent that extends ChatGPT’s capabilities with Codex-style task execution. The feature is designed for multi-step, longer projects rather than single prompts. OpenAI says ChatGPT Work gathers context from the apps, files, and workflows users select, can carry work forward for hours, and produces completed outputs such as documents, spreadsheets, presentations, and web applications.

The company is also rolling out GPT-5.6, a model family that includes Sol, Terra, and Luna. OpenAI presents Sol as its flagship model and says it delivers state-of-the-art performance in areas such as coding, knowledge work, cybersecurity, and science, while using fewer tokens. Terra and Luna are positioned as a balanced everyday work option and a more cost-efficient option, respectively. OpenAI states that GPT-5.6 includes safeguards such as real-time checks and monitoring, with particular emphasis on cybersecurity capabilities for tasks like code review and threat modeling.

OpenAI updates the desktop experience by merging Codex and ChatGPT into a unified application for Mac and Windows. Availability is tiered across plans: GPT-5.6 is rolling out broadly, while ChatGPT Work begins with Pro, Enterprise, and Edu before expanding to Plus and Business.