OpenAI unveiled three new AI models under the GPT-5.6 family — Sol, Terra, and Luna — but rolled them out only to a "small group of trusted partners" at the Trump administration's explicit request. Sol is the company's most capable flagship, designed for advanced agentic tasks with state-of-the-art performance in coding, biology, and cybersecurity, priced at $5 per million input tokens. Terra is a mid-tier model for everyday use at half the cost, while Luna is a faster, lower-cost option at $1 per million input tokens.
The White House cited safety concerns and a need to assess cybersecurity implications before broad public access — a move that follows heightened scrutiny of frontier AI releases since Anthropic's Fable 5 debut. OpenAI said the restricted preview would expand to the general public in the coming weeks.
OpenAI pushed back against making such restrictions permanent: "We don't believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." The company says it is working with the administration to build a repeatable safety-assessment framework for future model releases.