OpenAI launches ChatGPT Work, an agent built for hours-long project work

OpenAI has launched ChatGPT Work, a new agent inside ChatGPT designed to take on a goal, work independently across a user's connected apps and files for hours if needed, and hand back finished output rather than a chat reply.

What's new

From OpenAI's announcement: "It can gather information across your apps and workflows to create finished materials like sheets, slides, docs, and web apps, and stay with complex projects for hours by breaking them into smaller steps and completing them independently."

Rather than answering questions in a chat window, ChatGPT Work gathers context from connected apps and workflows, breaks a stated goal into smaller steps, and works through them independently, producing finished material — spreadsheets, slides, documents, and interactive web apps. OpenAI says users can follow its progress, answer questions, change direction, and approve important actions along the way, and that it can be chained across an entire workflow — for example, turning customer research into a campaign brief, then using that brief to generate and adapt marketing assets across markets. A Scheduled Tasks capability lets it keep projects moving while a user is away from their computer, including turning new Microsoft Teams and Slack messages into updated documents.

ChatGPT Work has "Codex technology built-in" and is powered by GPT-5.6, the model family OpenAI shipped the same day. OpenAI notes more than 5 million people already use Codex weekly, and more than 1 million of them now use it for work outside software development — context OpenAI cites for extending Codex's underlying technology into a general work agent.

Rollout starts today on web and mobile, beginning with ChatGPT Pro, Enterprise, and Edu plans, expanding to Plus and Business plans over the next few days. In the ChatGPT desktop app, Chat, Work, and Codex are available on every plan, including Free.

Context

ChatGPT Work lands the same day as the GPT-5.6 model family it runs on, and it's OpenAI's answer to a workplace-agent category that's been heating up all year. Anthropic shipped Claude Cowork earlier this year as its own persistent, multi-hour work agent, and Microsoft has been deepening GPT-5.6 integration into Microsoft 365 Copilot, announced the same day. OpenAI is positioning ChatGPT Work specifically around finished-artifact output — docs, sheets, slides, apps — rather than conversational assistance, and explicitly folding in Codex, previously marketed mainly as a coding agent, as core technology behind a broader work agent.

The staged rollout — Pro, Enterprise, and Edu first, Plus and Business to follow — mirrors how OpenAI has typically gated compute-intensive agentic features, prioritizing higher-tier and institutional accounts before wider consumer access.

Why it matters

The framing shift from "answer questions" to "go do the work and come back with a result" is the more consequential story than any single benchmark. If ChatGPT Work performs as described — genuinely staying on a multi-hour task and returning usable spreadsheets, slides, or working software rather than a chat transcript — it moves ChatGPT further into direct competition with project-management and knowledge-work software, not just other chatbots. Pairing the launch with GPT-5.6 on day one lets OpenAI market the new model family through a concrete, outcome-oriented use case rather than raw benchmark numbers, and folding Codex's adoption numbers into the pitch signals OpenAI sees its coding-agent traction as proof its agent technology generalizes beyond software work. It also sets up a direct, near-term comparison against Claude Cowork on the same kind of long-horizon, autonomous work.