OpenAI receives permission from the US government to release GPT-5.6 to the public this July 9

OpenAI will publicly launch this Thursday, July 9, the three variants of its new model family, GPT-5.6as announced by the company itself in

The reason for that initial limitation must be found in a cybersecurity order on artificial intelligence signed by Donald Trump at the beginning of June. The rule asks companies to voluntarily submit their most powerful models for a government review thirty days before their public launch. In the case of GPT-5.6, this translated into an initial deployment restricted to government-approved entities only. OpenAI went on to state at the time that it does not believe this type of government access process should become the norm in the long term, although it also acknowledged that meeting the requirement was the quickest route to bringing its latest family of models to the general public.

Clear passage before the expected thirty days

Finally, OpenAI will not have to wait the full month required by the executive order to expand access to GPT-5.6. As reported by the American media Axios, Trump administration granted company permission for broader deployment after subjecting the model to additional testing and having several follow-up meetings with OpenAI. This verification process was carried out by the Center for Standards and Innovation in AI of the Department of Commerce, and OpenAI itself sent technical experts to Washington to resolve the government’s doubts and concerns as quickly as possible.

This move confirms that, at least in practice, the thirty-day period set by the cybersecurity order can be substantially shortened when the developer actively collaborates with government testing, something that sets a relevant precedent for the rest of the artificial intelligence companies subject to the same regulations.

Three variants with very different prices

The GPT-5.6 family is It consists of three different models for performance and cost. Sun It is, according to OpenAI itself, the most powerful model that the company has developed to date. Terradesigned for everyday use, promises similar performance to GPT-5.5 despite costing half as much, while Moon It is positioned as the most economical option of the three.

In terms of prices, Sol costs 5 dollars (about 4.3 euros at the current exchange rate) for each million input tokens and 30 dollars (about 25.8 euros) for each million output tokens. Terra stands at $2.50 (approximately 2.15 euros) per million tokens for entry and $15 (approximately 12.9 euros) for exit. Luna, the most affordable variant, costs $1 (0.86 euros) per million tokens in and $6 (5.16 euros) per million tokens out. It is worth remembering that These figures correspond to the price of the API in the United States.

The case of Anthropic with Mythos and Fable

The same cybersecurity order also affected Anthropic, which had to completely block access to its most recent models, Mythos and Fablesto comply with the government’s requirement to prevent users from outside the United States from using them.

The parallel between both cases, that of OpenAI and that of Anthropic, makes it clear that the executive order on cybersecurity in AI is having a direct and measurable impact on the launch schedules of the large artificial intelligence laboratories, forcing all of them to intersperse government review processes within their usual commercial roadmaps.

For companies already integrating OpenAI models into their workflows, the arrival of GPT-5.6 means access to a wider range of cost and performance optionswith Terra as a particularly attractive alternative by offering results close to the previous generation for half the price. The simultaneous availability of the three variants, Sol, Terra and Luna, also allows the spending on tokens to be adjusted according to the type of task, reserving the most powerful model for cases that really require it and diverting the rest of the traffic towards cheaper options.

The episode also shows that, in the regulatory field, generative artificial intelligence has come to occupy a central place on the agenda of the US administration, with review processes that already directly condition when and how the most advanced models from companies such as OpenAI or Anthropic reach the market.