OpenAI was cleared to widely launch its most advanced model, after the US government approved a broader launch of GPT-5.6 that had been delayed for weeks under Washington’s new oversight regime for frontier AI.
Until now the model had only been available through a preview restricted to about 20 members whose names were individually approved by the United States government. That deal, the first of its kind for a U.S. Frontier model, is what now replaces the broader rollout.
The approval came after additional testing by the Commerce Department’s AI Standards and Innovation Center, the body created to review advanced systems. OpenAI sent technical experts to Washington to answer the agency’s questions, according to Axios.
GPT-5.6 is a three-tier family rather than a single model. Sol is the flagship, Terra is a lower-cost mid-tier option, and Luna is the fastest and cheapest of the three.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and some questionable AI art. It’s free, every week, in your inbox. Register now!
OpenAI described Sol as strong in coding, biology, and cybersecurity, and paired him with a “maximum reasoning effort” mode that gives the model more time to solve difficult problems. Those same capabilities, particularly in biology and cybernetics, are part of the reason the government wanted a closer look before broad disclosure.
Tiering is both a commercial and technical choice. Terra is aimed at everyday enterprise workloads where cost matters more than raw capacity, while Luna is designed for high-volume tasks that need speed above all else, a division that allows OpenAI to charge very different prices on the same family.
The preview it now replaces was unusually tight. For weeks, GPT-5.6 was available only to a short list of organizations whose identities OpenAI had shared with the government, the first time a US lab had closed a frontier model behind a state-approved list.
The review falls within a framework the Trump administration established on June 2, which introduced voluntary pre-launch verification for the most capable models. The GPT-5.6 case went further, moving from a voluntary review to a government-run access list, a step that OpenAI had taken. He agreed only after being asked to delay the launch..
OpenAI has made it clear that it is uncomfortable with the precedent. The company said it does not believe this type of government access process should become the default in the long term, although it agreed to participate this time.
The discomfort is not difficult to understand. A government that can block a launch can also stop it, a power the administration has already used elsewhere in the sector by order Anthropic to close two models.
For OpenAI, the business risks of the delay were real. Every week that GPT-5.6 remained within a 20-partner preview was a week in which rivals could court the enterprise customers they wanted to reach with the new levels.
The company now hopes to expand access to GPT-5.6 within days, building on the foundation it laid with GPT-5.5 at the beginning of the year. It has said all three tiers will be widely available in the coming weeks, although it has not set a public date.
OpenAI is not the only laboratory within the new regime. The same framework covers its rivals, meaning that how this launch has played out will likely determine how any American company’s next frontier model reaches the public.
What the episode establishes is less a product timeline than a template. For the first time, a major U.S. lab has released a border model based on the government’s timeline rather than its own, and both sides now have to decide whether this is a one-time model or the shape of things to come.
