OpenAI began a limited preview of its GPT-5.6 series, comprising Sol, Terra and Luna, on Friday, releasing the models only to partners individually approved by the Trump administration. According to Axios, it’s the first time the U.S. government has preemptively asked an American AI company to restrict a model launch.

The pressure came from the White House’s Office of the National Cyber Director and Office of Science and Technology Policy, which cited security concerns, per Axios. Bloomberg reported the administration is clearing customers one by one during the preview window. Sam Altman discussed the model with Commerce Secretary Howard Lutnick on Wednesday.

OpenAI didn’t hide its discomfort. “We don’t believe this kind of government access process should become the long-term default,” the company said, while planning general availability in the coming weeks.

The timing isn’t accidental. Under an executive order, the administration must stand up a classified process by August to assess AI cyber capabilities and designate “covered frontier models.” Friday’s preview is effectively a dry run for that regime, conducted before the statutory machinery exists.

OpenAI’s own framing emphasizes capability without crossing a line: Sol is its most capable model to date, with agentic gains in coding, biology and cybersecurity, new “max” and “ultra” reasoning modes, and performance competitive with Anthropic’s Mythos Preview on ExploitBench at roughly a third of the output tokens. The company says Sol doesn’t cross its “critical” cybersecurity risk threshold.

Pricing, per TechCrunch, lands aggressively: Sol at $5 per million input tokens and $30 per million output tokens, Terra at half that ($2.50 and $15), and Luna at $1 and $6.

The juxtaposition is the story. A model deemed below the critical-risk bar still required White House sign-off on every customer. The export-control logic of semiconductors has quietly arrived at the API layer.

Sources