Enterprise

OpenAI ships ChatGPT Work agent and GPT-5.6 model family in enterprise blitz

The company released Sol, Terra and Luna alongside a new long-horizon work agent, ending a government-imposed delay on the launch.

By Asher Glick · Enterprise software desk · 2026-07-10 · 3 min read

Photo: Unsplash / Mohammad Rahmani — Laptop screen showing lines of code, illustrating AI-assisted software development

OpenAI on Thursday launched ChatGPT Work, a long-horizon agent designed to field business tasks for hours at a stretch, alongside a three-tier GPT-5.6 model family branded Sol, Terra and Luna. The bundle arrives after what Sam Altman described as “many changes” following a “collaborative back and forth” with the Trump administration, which had asked OpenAI to limit the initial rollout to “a small group of trusted partners.”

The pricing tells you where the fight is. Sol runs $5 per million input tokens and $30 output. Terra sits at $2.50 and $15. Luna comes in at $1 and $6. Three price points, three deployment postures, one message to procurement departments already sharpening their pencils on AI spend.

“Every enterprise now is thinking about spend and the value they’re getting in exchange for AI, and this is what we really want to do,” Altman told CNBC’s Julia Boorstin. He also called the administration’s technical capabilities “impressive,” a phrasing that reads as narrative management aimed at both regulators and the enterprise buyers watching the exchange.

OpenAI positions Sol as its best coding model yet, and the company’s own claim is unusually numerical: Sol “sets a new state of the art at 80, 2.8 points above Fable 5, while using less than half the output tokens, taking less than half the time, and costing about one-third less” on the Artificial Analysis Coding Agent Index. Altman said Sol is “as good or better” than competing models on the market, and OpenAI reports it’s 54% more token efficient on agentic coding tasks. Availability spans ChatGPT, Codex and the API.

ChatGPT Work drops OpenAI directly into Microsoft Copilot’s lane, in the same week Meta unveiled Muse Spark 1.1 and SpaceX-owned xAI shipped Grok 4.5 on Wednesday. The clustering isn’t coincidence. Both OpenAI and Anthropic have confidentially filed IPO prospectuses, SpaceX absorbed xAI earlier this year, and every serious lab is now selling agents rather than chat.

The token-efficiency framing is the giveaway. When a frontier lab starts leading with cost-per-task rather than capability curves, the market has moved from demo to line item. That’s not a plateau. It’s an industry learning to sell to a CFO.

Sources