OpenAI has received approval for the broad release of its most advanced model, after the US government signed off on a wider rollout of GPT-5.6, which had been delayed for weeks under Washington's new oversight regime for AI pioneers.
See also: OpenAI releases GPT-5.2 for ChatGPT users
Until now, the model has only been available through a limited preview to about 20 partners, whose names have been individually approved by the US government. This arrangement, the first of its kind for a US pioneering model, is what now goes beyond wider release. The approval followed additional testing by the Commerce Department’s AI Standards and Innovation Center, which was established to evaluate advanced systems. OpenAI sent technical experts to Washington to answer the agency’s questions.
GPT-5.6 is a family of three tiers, not a single model. Sol is the top tier, Terra a more affordable middle ground, and Luna the fastest and cheapest of the three. OpenAI has described Sol as strong in coding, biology, and cybersecurity, and has paired it with a “maximum effort logic” mode that gives the model more time to tackle tough problems. Those same capabilities, particularly in biology and cybersecurity, are part of the reason the government wanted a closer look before wide release.
The scaling is as much a commercial choice as it is a technical one. Terra is aimed at everyday business tasks where cost matters more than raw capacity, while Luna is built for high-volume tasks that need speed above all else, a division that allows OpenAI to charge very different prices within the same family.
See also: GPT-5.5-Cyber: OpenAI strengthens Daybreak for code security
The preview that now goes beyond was unusually strict. For weeks, GPT-5.6 was available only to a small list of organizations whose identities had been shared by OpenAI with the government, the first time a U.S. lab had restricted a pioneering model behind a government-approved list.
The review is part of a framework established by the Trump administration on June 2 that introduced a voluntary pre-release review for the most capable models. The case of GPT-5.6 went further than that, moving from voluntary review to a government-managed access list, a step OpenAI had agreed to take only after a request to delay the release.
OpenAI has made it clear that it is not comfortable with the precedent. The company said it does not believe this government access process should become the long-term default, while it has agreed to participate this time. The discomfort is not hard to understand. A government that can restrict a release can also stop it, a power that the administration has already used elsewhere in the sector by ordering Anthropic to shut down two models. For OpenAI, the commercial risks of the delay were real. Every week that GPT-5.6 remained in preview with 20 partners was a week that competitors could attract the enterprise customers it wanted to reach with the new levels.
The company now expects to expand access to GPT-5.6 within days, building on the foundation it created with GPT-5.5 earlier this year. It has said that all three levels will become generally available in the coming weeks, although it has not set a public date. OpenAI is not the only lab under the new regime. The same framework applies to its competitors, meaning that how this release progresses is likely to shape how the next leading-edge model from any US company reaches the public.
See also: Google Gemini co-head Noam Shazeer leaves for OpenAI
What the episode establishes is less a product timeline and more a pattern. For the first time, a leading American lab has released a pioneering model on the government’s timeline rather than its own, and both sides must now decide whether this was a one-off or the shape of things to come.
