OpenAI has started a limited preview of its new GPT-5.6 model family: Sol, Terra and Luna. The company said it previewed its launch plans and model capabilities to the U.S. government in advance, then began with a small group of trusted partners at the administration’s request. OpenAI plans to make the models generally available in the coming weeks, but has not provided an exact date.

The flagship model is GPT-5.6 Sol, which OpenAI describes as its strongest model yet. It shows improved capabilities in coding, biology workflows and cybersecurity. Those stronger cyber capabilities are also why the rollout is cautious. In its system card, OpenAI classifies Sol, Terra and Luna as High capability for cybersecurity and biological/chemical risk, while still below the Critical threshold.

OpenAI says it does not believe this kind of government access process should become the long-term default. The company acknowledges that it delays access for users, developers, enterprises and cyber defenders. Still, it frames the limited preview as a temporary step toward broader availability and safer deployment.

The broader context is regulation. OpenAI has been calling for a federal framework for frontier AI, including a stronger role for CAISI in evaluating the most capable models. The GPT-5.6 rollout shows that releasing frontier AI is becoming not only a technical decision, but also a political and national-security process.