OpenAI launches GPT-5.6 Sol, Terra, and Luna after government-supervised preview
The three-tier frontier family went to general availability July 9 following a Commerce Department review that gated its release, alongside a new long-running enterprise agent called ChatGPT Work.
OpenAI pushed its GPT-5.6 family, Sol, Terra, and Luna, to general availability on Thursday, closing out a limited preview that the Trump administration had gated for weeks before the Commerce Department cleared broader distribution. The three-tier structure is the first frontier release to ship after a formal U.S. government review cycle, and the framing around that review is already contested.
Pricing sets the hierarchy plainly. Sol lists at $5 per million input tokens and $30 per million output; Terra at $2.50 and $15; Luna, the cost-optimized variant, at $1 and $6. All three are live across ChatGPT, Codex, and the OpenAI API. A new “ultra” setting lets Sol coordinate multiple agents across parallel workstreams, with Axios reporting that the mode allows delegation to submodels.
The benchmark sheet is aggressive. Sol posts 53.6 on Agents’ Last Exam, a 55-field evaluation of long-running professional workflows, 13.1 points ahead of Anthropic’s Claude Fable 5. It claims state-of-the-art on BrowseComp at 92.2% and OSWorld 2.0 at 62.6%, and an 80 on the Artificial Analysis Coding Agent Index. OpenAI says Sol uses 85% fewer output tokens than Anthropic’s Opus 4.8 on OSWorld; Sam Altman has separately claimed a 54% token-efficiency gain over prior versions on agentic coding tasks.
Then there’s the regulatory choreography. Axios reported last month that Washington had limited initial access to government-approved entities, and CNBC’s Axios-sourced report this week said Commerce cleared the wider rollout. The White House told CNBC it had given OpenAI no “green light, approval or clearance.” OpenAI’s launch post threads the needle, saying it had “previewed our plans and the models’ capabilities” with the U.S. government.
Alongside the models, OpenAI unveiled ChatGPT Work, a long-running agent designed to handle multi-hour tasks and produce documents, spreadsheets, presentations, and web applications. Per Axios, it pulls context from connected apps and files, shipping first on Mac and Windows with web to follow. Sol will also be served on Cerebras hardware at up to 750 tokens per second beginning this month.
The pattern echoes the 2023 White House voluntary commitments, only now the preview itself is the policy instrument. Approval isn’t declared; it’s staged.
Sources
- https://openai.com/index/gpt-5-6/
- https://techcrunch.com/2026/07/09/openai-launches-its-new-family-of-models-with-gpt-5-6/
- https://www.bloomberg.com/news/articles/2026-07-09/openai-unveils-chatgpt-work-agent-to-field-tasks-for-hours
- https://www.axios.com/2026/07/09/ai-openai-gpt-release
- https://www.cnbc.com/2026/07/08/openai-gets-us-regulatory-approval-for-gpt-5point6-rollout-axios-report.html
— END —