OpenAI has launched three new AI models called Sol, Terra, and Luna, but their release has not been made globally available in the usual way. Instead, the company said the models will initially be accessible only to a small group of approved partners after a request from the US government for a restricted rollout.
The announcement was made by OpenAI CEO Sam Altman on X, where he described the models as a significant improvement in capability and efficiency, especially in terms of pricing. However, he also confirmed that full public access is being delayed due to regulatory requirements involving early safety review of advanced AI systems.
Altman said Sol is an upgraded, efficient model offering stronger performance while being priced similarly to GPT-5.5. Terra, part of the GPT-5.6 lineup, is designed to match GPT-5.5 performance but at roughly half the cost.
A third model, Luna, has also been introduced under the same GPT-5.6 umbrella, with all three positioned as part of OpenAI’s next-generation development focused on better performance and lower operational cost for businesses and developers.
Responding to a user query about global availability, Altman said the company is “working hard for worldwide” access.
According to OpenAI, the models are currently being offered in a limited preview to selected partners following a request from the US government. The company also stated that it had already shared model details with regulators before launch as part of early evaluation procedures.
OpenAI added that while it supports wider access, the current restricted rollout should not become a permanent model, as it could slow down access for developers, enterprises, and security teams who rely on such tools.
The company also said it is collaborating with US authorities to develop a more structured and repeatable process for reviewing advanced AI systems before release, aiming to balance safety oversight with faster deployment.
This development comes amid increasing regulatory attention on frontier AI models in the United States, with authorities pushing for earlier review and tighter oversight before public release.
OpenAI said Sol shows notable improvements in areas such as coding, biology, and cybersecurity-related tasks, particularly in identifying and fixing software vulnerabilities, while still operating within its safety constraints.
