GPT-5.6 officially launches, expanding to ChatGPT Work and Copilot

What was newly announced?

In an official post on July 9, OpenAI said it is making the GPT-5.6 family of models generally available. The Sol, Terra, and Luna lineup, first introduced through a limited preview in late June, is now entering a broader product flow.

According to OpenAI, Sol is the flagship model aimed at the highest reasoning limits, Terra is a balanced model for everyday work and agent tasks, and Luna is a lighter model focused on faster, more cost-efficient use. This news is not only about a “new model performance chart”; it also shows which work tools the same model family will be attached to.

What does ChatGPT Work change?

On the same day, OpenAI also introduced ChatGPT Work. In the official description, ChatGPT Work is not simply a chatbot that answers questions, but an agent that refers to apps, files, and workflows to create outputs such as sheets, slides, documents, and web apps.

Put simply, it is an attempt to push AI from “an AI that writes answers” into “a work partner that breaks tasks into multiple steps and produces actual deliverables.” OpenAI explains that Codex technology is included and that complex projects can be divided into smaller steps and carried forward over several hours. For everyday users, this may look like AI moving beyond search, organization, and drafting toward handling actual bundles of work.

It also extends to Microsoft 365 and GitHub Copilot

OpenAI said GPT-5.6 is becoming the preferred model in Microsoft 365 Copilot. The announced scope includes Word, Excel, PowerPoint, Chat, and Cowork. This points to a flow in which the new model family is used inside workplace tools for writing documents, analyzing data, and organizing presentations.

GitHub also announced through its Copilot Changelog on the same day that GPT-5.6 Sol, Terra, and Luna are being rolled out to Copilot in phases. Based on GitHub’s explanation, Sol is aimed at large codebases and long-running agent work, Terra at everyday coding, and Luna at smaller, faster tasks. This makes the structure for matching model choice to coding cost and task difficulty clearer.

How this differs from the June limited preview

Game Sunakku already covered the limited preview of GPT-5.6 Sol in late June. At that time, the key points were the new model family, access restrictions, safety review, and pricing structure. The July 9 news is the next step. Models introduced through the limited preview have expanded into real product touchpoints such as ChatGPT Work, Microsoft 365 Copilot, and GitHub Copilot.

So this article is not repeating the same announcement. It is a follow-up that organizes the movement of GPT-5.6 from a “model that can be used” to a “default option inside work tools”.

What users should watch carefully

That said, as agents become more powerful, there is also more to check immediately. Features like ChatGPT Work that move across apps and files need to be considered together with permission scope, company data policies, log retention, and responsibility for reviewing outputs. On the GitHub Copilot side, model-based usage billing and administrator policies are also important.

In short, this GPT-5.6 follow-up is not only about competition over model performance. It is also about a change in how workplace tools and coding tools are operated. When choosing AI tools from now on, users need to look not only at “how smart is it?” but also at “how far can it access, how much will it cost, and who reviews the result?”.