ChatGPT Work: OpenAI’s New AI Agent Turns Goals Into Finished Work
OpenAI has launched ChatGPT Work, a new agent built into ChatGPT that completes multi-step tasks across a person’s connected apps and files. Rather than answering one question at a time, the tool takes a single goal and works through it independently — producing documents, spreadsheets, presentations, and reports.
Powered by Codex and the newly released GPT-5.6 model, ChatGPT Work began rolling out this week, starting with Pro, Enterprise, and Edu subscribers before expanding to Plus and Business users.
What Is ChatGPT Work?
ChatGPT Work is an autonomous agent inside ChatGPT designed to carry out entire workflows instead of responding to isolated prompts. A person can describe a broader goal, and the agent pulls context from selected apps and files to complete it end to end.
OpenAI says the agent can gather information across a user’s apps and workflows to produce finished materials such as sheets, slides, and documents, and can stay with complex projects for hours by breaking them into smaller steps it completes on its own.
The agent is built on Codex, OpenAI’s coding-focused system, which the company had already expanded beyond software development to support general productivity work.
Powered by GPT-5.6
ChatGPT Work runs on GPT-5.6, a new model family released alongside it. GPT-5.6 comes in three tiers — Sol, Terra, and Luna — representing different tradeoffs between intelligence, speed, and cost. Sol is the most capable version, Terra is built for everyday balanced use, and Luna is optimized for speed and affordability.
OpenAI describes Sol as delivering strong reasoning across multi-step tasks and generating output that matches a user’s own templates and reference material. An “ultra” mode within Sol can also coordinate multiple subagents for particularly demanding jobs.
How It Works
ChatGPT Work isn’t limited to generating text. The agent can draw on local files and apps, or use a built-in browser to reach websites and online tools. Plugins allow people to connect external apps and systems directly to it, and a new “Sites” feature, currently in beta, lets users build interactive reports, live dashboards, and simple web apps.
Tasks can be started on a phone, left running in the background on a desktop, and checked from any device — a setup meant to let people launch work and return to it later instead of waiting on a single reply.
Why It Matters
The launch reflects a broader shift in how AI companies want chatbots to be used — not just for answers, but for finished output that requires little follow-up. Tools like this target people who aren’t professional coders but want the capability of coding-based agents, which often outperform standard chatbots yet typically require specialized skills to operate.
It also raises the stakes in the enterprise AI market, where major AI companies are competing to build agents capable of handling complex workplace tasks with minimal human input. Pricing efficiency is part of that pitch: OpenAI has said its Sol model is meaningfully more token-efficient on agentic coding tasks, a detail aimed at businesses increasingly focused on the cost of running AI tools at scale.
Key Features
- Cross-app task execution — completes goals that span multiple connected apps and files rather than a single input
- Autonomous multi-step workflows — breaks large projects into smaller steps and works through them independently over time
- Document and report generation — produces spreadsheets, slides, documents, and web-based reports styled to a user’s own templates
- Browser and plugin support — pulls data from external websites and connected third-party systems
- Cross-device task management — tasks can be started on mobile and monitored or finished on desktop
- Unified desktop app — Codex and ChatGPT have been merged into a single desktop application for Mac and Windows
Who Gets Access?
Access is rolling out in stages:
- Pro, Enterprise, and Edu users — the first to receive ChatGPT Work, on web and mobile
- Plus and Business users — access is expanding to these plans over the following days
- Desktop users, including Free accounts — the updated desktop app for Mac and Windows, which includes Chat, Work, and Codex, is available to everyone
- Existing Codex app users — can simply update their app, which becomes the new unified desktop client
Within paid tiers, some capabilities are gated further. The highest-capability “ultra” mode, for example, is limited to Pro and Enterprise users in ChatGPT Work.
Conclusion
ChatGPT Work represents a clear step toward AI tools that finish real work rather than just respond to questions. How well it performs unsupervised, over longer and more complex tasks, will likely determine how quickly it’s adopted — particularly as competition among AI companies building similar workplace agents continues to intensify. For now, access begins with paid tiers, with broader availability expected in the coming weeks.
Frequently Asked Questions
What is ChatGPT Work?
An AI agent inside ChatGPT that completes multi-step tasks across connected apps and files, turning a single request into finished documents, spreadsheets, or reports.
What model powers ChatGPT Work?
GPT-5.6, OpenAI’s newest model family, available in three tiers: Sol, Terra, and Luna.
Who can use ChatGPT Work right now?
Pro, Enterprise, and Edu users have access first, with Plus and Business following shortly. The updated desktop app is available to all users, including Free accounts.
How is ChatGPT Work different from standard ChatGPT?
Standard ChatGPT responds to individual prompts, while ChatGPT Work can independently work through multi-step projects over extended periods using context from a user’s apps and files.
Does ChatGPT Work require coding skills?
No. It’s designed to give non-coders access to the capabilities of coding-based agents, which typically require specialized skills to use directly.