OpenAI has announced the public launch of a new line of flagship GPT-5.6 models — Sol, Terra, and Luna. The models’ general availability was delayed by more than a month due to the Trump administration’s requirement to test powerful AI models for security before release.
OpenAI has announced the public launch of a new line of flagship GPT-5.6 models — Sol, Terra, and Luna. The models’ general availability was delayed by more than a month due to the Trump administration’s requirement to test powerful AI models for security before release.
In early June, President Trump signed an executive order on AI cybersecurity that requires companies to submit powerful models for government review at least 30 days before public release. The review of GPT-5.6 was conducted by the U.S. Department of Commerce’s Center for AI Standards and Innovation (CAIS); OpenAI even sent technical experts to Washington to answer questions during the review.
Initially, access to the models was limited to a limited number of trusted partners and government agencies approved by the government. OpenAI emphasized that it did not consider this format of government access to be the norm in the long term, but agreed to it to accelerate the release of the models. After additional testing and consultation, the Trump administration authorized a wider release.
Interestingly, almost simultaneously, due to similar export restrictions by the US government, Anthropic also temporarily suspended access to its most powerful models — Mythos and Fable; it has already received permission to return Mythos 5 and plans to restore access to Fable 5 later.
The model line consists of three options for different tasks:
GPT-5.6 Sol, released on July 9, has significant improvements in programming, complex logical reasoning, cybersecurity, and scientific tasks. The model also has Max and Ultra modes, which use more computing resources or multiple subagents to solve complex tasks.
The company also reports that the new generation of models showed the best results in internal tests for writing code and executing complex agent scripts, and also received enhanced security mechanisms.
API prices: Sol costs $5 per million incoming tokens and $30 per million outgoing; Terra — $2.5 per incoming and $15 per outgoing; Luna — $1 per incoming and $6 per outgoing tokens.
The rollout is happening in waves: access via the Enterprise API opened on July 7, ChatGPT Plus subscribers will receive the model on July 10, and the full rollout on the Team tariff is expected by mid-July.
