OpenAI Rolls Out GPT-5.6 After Government Greenlight, Debuts ChatGPT Work

OpenAI has opened its GPT-5.6 model to the general public after receiving the green light from the Trump administration. For about two weeks, according to The Verge, the model had been caught up in a regulatory tangle: it was rolled out only to government-approved organizations during what the company called a "limited preview" period. That period, the company says, is now over.

OpenAI CEO Sam Altman called GPT-5.6 "the best model we have ever produced." To mark the occasion, the company also unveiled a new AI agent the same day: ChatGPT Work.

What ChatGPT Work Is

ChatGPT Work is billed as a combination of ChatGPT and Codex — OpenAI's coding-focused tool — designed to let non-technical users take advantage of Codex's capabilities for tasks that have nothing to do with writing code. It runs on the GPT-5.6 model suite, made up of three pieces OpenAI has named Sol, Terra, and Luna.

As the company explained in a blog post, the agent "can gather context from the apps, files, and workflows you choose and create finished materials such as documents, spreadsheets, presentations, and web apps." What makes this possible is what OpenAI calls a "unified plugins directory," which lets ChatGPT connect to tools like Slack, Gmail, Google Drive, calendars, and CRMs.

In practice, the idea is that users won't have to jump between apps or figure out how things fit together behind the scenes. They ask for an output — a report, a spreadsheet, a presentation — and the agent pulls the necessary information from authorized sources and delivers the finished work.

Who Can Use It, and When

The rollout is staggered. Mac and Windows users worldwide, including free ChatGPT users, should have immediate access to both ChatGPT Work and GPT-5.6 through the desktop app.

On mobile and the web, the order is different: Pro, Enterprise, and Edu users get access first, while Plus and Business users will receive access "over the next few days," according to OpenAI. The company added that the "rollout is starting globally and will continue gradually toward full availability over the next 24 hours."

The Race to Build an Agent Useful for Everyone

The move fits into a competition that's been simmering for a while. Companies like OpenAI and Anthropic, alongside tech giants such as Google — and, more recently, Apple — have spent months trying to make AI agents genuinely useful for the average person. The results, as The Verge notes, have been mixed: an AI agent that acts as a reliable right hand for the everyday consumer remains, for now, out of reach.

That push has also been fueled by the emergence of OpenClaw, an open-source AI agent that went viral. The pressure, then, isn't coming only from the big labs — it's also coming from open projects showing what's possible outside the structures of major companies.

ChatGPT Work is a direct competitor to Anthropic's Claude Cowork, which follows the same logic of merging its Claude chatbot with its Claude Code programming tool. Both bets start from the same intuition: the capabilities developed for coding — breaking down a task, executing steps, handling external tools — are useful for much more than writing software, and whoever puts them within reach of non-technical users will have an edge.

Sol and the Cost Question

Within the GPT-5.6 suite, OpenAI is especially banking on Sol, the most powerful of the three models. The company is positioning it to set "a new standard for intelligence and efficiency," particularly when it comes to coding, cybersecurity, and science, as well as computer-use capabilities — that is, the model's ability to operate a computer the way a person would.

There's a second selling point, and it's economic. OpenAI is marketing the model as a lower-cost alternative to competitors' most powerful models. That message lands at a moment when the industry is grappling with complaints about a broader financial squeeze and about how AI lab costs keep getting passed on to customers. Training and running these models is expensive, and that bill tends to trickle down to the end user. Presenting Sol as a lower-cost option is, in that context, as much a commercial decision as a technical one.

Why the Government Sign-Off Matters

The most striking detail in this announcement isn't the product itself — it's the approval process that preceded it. The fact that a language model had to go through a "limited preview" period restricted to government-approved organizations before it could be opened to the public signals a kind of relationship between AI labs and political power that wasn't the norm during earlier waves of these products, when a public launch was simply the default path.

The available reporting doesn't detail the specific reasons behind that initial restriction or the terms of the approval. But the fact itself — that a public release depends on official sign-off — suggests that the line between launching a model and getting permission to launch it is narrowing, at least for the most powerful systems.

For users, the immediate effect is straightforward: GPT-5.6 and ChatGPT Work are already reaching their devices, starting with desktop and expanding by plan tier over the coming hours. The bigger question — whether this generation of agents finally delivers on the promise of handling office work start to finish, or falls short again like earlier attempts — will have to be answered by real-world use, not by the announcement itself.