OpenAI Debuts GPT-5.6 Models and ChatGPT Work Agent

OpenAI has now made GPT-5.6 available to everyone while simultaneously introducing ChatGPT Work, which generates complete documents and files. The announcement arrives two days after Anthropic broadened access to its rival product.

OpenAI rolls out GPT-5.6 family of models. Sam Altman called the new system the strongest his firm has built to date. GPT-5.6 actually comprises three separate models renamed under a fresh scheme in which the numeral marks the generation and Sol, Terra plus Luna reflect tiers of capability. Sol serves as the top-tier option focused on programming, cybersecurity and scientific work. Terra handles large-scale corporate workloads, while Luna delivers quicker and cheaper performance.

For two weeks the models stayed limited to roughly twenty hand-picked groups under a preview cleared by Washington. The release reportedly needed approval from the Trump administration before wider distribution could begin.

Pricing and availability details for developers. Sol carries a reported price of $5 per million input tokens and $30 per million output tokens. Luna comes in lower at a reported $1 per million input tokens and $6 per million output tokens. A token amounts to a piece of a word, so longer inputs raise the final cost.

The models have begun appearing inside ChatGPT, Codex and the API. The tweet from OpenAI on July 9, 2026 stated: "Sol, Terra, and Luna, our GPT‑5.6 family of models, are starting to roll out now in ChatGPT, Codex, and the API."

ChatGPT Work agent targets non-technical users. ChatGPT Work extends Codex-like abilities to everyday employees who avoid coding. It draws information from connected apps and stored files before creating reports, spreadsheets or slide decks. A plugin directory links the service to Slack, Gmail, Google Drive, calendars and customer relationship management software. Free users on Mac and Windows computers receive instant access, whereas mobile and web versions arrive gradually.

Timing signals direct competition with Anthropic. Anthropic first launched Claude Cowork for desktop computers in January and expanded it to phones and browsers shortly before OpenAI's July 9 reveal. Each firm offers an agent built on its core model that runs quietly, sends mobile alerts for choices, and awaits explicit permission before acting. Both target office employees who avoid command lines. The most telling data comes from Anthropic out of 1.2 million Cowork sessions.