OpenAI announces GPT-5.6 Sol, limits rollout to trusted partners per US gov request

OpenAI on Friday announced three new AI models—GPT-5.6 Sol, Terra, and Luna—tiered by capability, with Sol being its strongest offering yet. The company said it is complying with a U.S. government request to initially limit the rollout to a 'small group of trusted partners' and plans to make the models generally available within weeks. OpenAI did not disclose which partners can use the models.

Sol shows improvements across coding and biology and is OpenAI's most capable model for cybersecurity, though it still doesn't cross into the company's 'critical' risk threshold defined as bringing 'unprecedented new pathways to severe harm.' OpenAI said it previewed the models' capabilities with the government ahead of Friday's launch and is working with the Trump administration to establish a framework for voluntary model capability assessments. The company stated it is building a 'repeatable process for future model releases.'

The announcement comes two weeks after rival Anthropic disabled access to two of its latest models to comply with an export control directive from the Trump administration. Anthropic remains in active negotiations with officials but has not said when its models will return. The Trump Administration's AI executive order, signed earlier this month, asked developers to voluntarily allow government model assessment ahead of full release, though it offered few specifics.

For operators: government pre-release review of frontier models is now the de facto process. Both Anthropic and OpenAI are navigating the same friction point—trusted-partner gating delays public availability and creates asymmetric market access. The question is whether this becomes standard practice or whether the phased access creates supply-chain leverage for the government while competitors fragment user bases by region and trust status.