OpenAI has now begun offering its artificial intelligence model to a limited number of select partners – approved by the US government. The GPT-5.6 model comes in three different variants, each known as Sol, Terra and Luna.
Sol is the most powerful model including for reasoning, Terra is for performance equivalent to GPT-5.5 but at a lower price, and finally Luna is their low-cost model. GPT-5.6 is also said to be optimized for the era of agent-based AI, including for various tasks that require reasoning, and deep thinking.
It is said that OpenAI took the step of offering it in a limited way due to the directive of the US government, which made them not offer it widely. Previously, Anthropic had offered Fable 5, but was directed by the US government to order the model to be offered only to their citizens, and banned for foreign citizens – including even Anthropic employees.
OpenAI says they have added several additional features focused on safety and security, including preventing their models from being jailbroken.
It is currently unknown when OpenAI will offer this model publicly to everyone. In their statement, OpenAI also emphasized that they hope that this government-involved access process should not become the norm moving forward.
