OpenAI will restrict the initial release of its new model lineup to a small group of vetted partners after the US government asked the company to start with a limited preview. The decision gives federal agencies, defense contractors, and cybersecurity teams a first look at the technology before it reaches the broader public.
Three new models, one flagship
The rollout includes Sol, OpenAI's most capable model to date, which the company says has improved agentic abilities in coding, biology, and cybersecurity. Two additional models-Terra, a general work assistant, and Luna, described as a fast, lower-cost option-complete the lineup. Full public availability is expected in the coming weeks.
For government professionals who will eventually work with these tools, AI for Government training provides context on how such models intersect with public-sector missions. OpenAI Courses cover the company's technology in depth, from underlying architecture to practical use cases.
Limited preview at the government's request
"As part of our ongoing engagement with the U.S. government, we previewed our plans and the models' capabilities ahead of today's launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly," OpenAI said in a blog post.
The company did not name the selected partners. It described the arrangement as a short-term step while it continues testing and coordination.
A short-term step, not a long-term default
OpenAI warned that government-moderated access should not become permanent. "We don't believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them," the company said.
The move is the first public implementation of the voluntary AI review process outlined in a recent Trump administration executive order on frontier models. That framework lets developers submit covered models for government review for up to 30 days before a wider release. The same order has already affected other AI firms: in June, Anthropic disabled foreign access to two models after an export control directive from the US government.
Why this matters for Government
For officials across federal and state agencies, the limited preview signals that frontier AI models are now a direct component of national security and technology policy. Early access offers a chance to assess capabilities, shape use policies, and plan procurement before public release. The arrangement also highlights that government coordination with AI developers is becoming routine-a shift that demands technical literacy and clear governance frameworks inside the public sector.
Your membership also unlocks:
