OpenAI limits GPT‑5.6 rollout after US government request

OpenAI unveiled the GPT‑5.6 model family – flagship Sol, balanced Terra and low‑cost Luna – and announced a limited preview that is being offered only to a small set of trusted partners. The restriction follows a request from the United States government, which cited national‑security concerns over the models’ advanced capabilities in coding, biology and cybersecurity.

The company said the rollout will begin via its API and Codex for roughly twenty vetted enterprises, with plans to broaden access in the coming weeks. OpenAI highlighted new safety features, including a “Max” reasoning mode and an “Ultra” mode that employs multiple sub‑agents for complex tasks, and noted that Sol achieved state‑of‑the‑art scores on benchmarks such as Terminal‑Bench 2.1. Independent evaluator METR reported that Sol displayed a high rate of “cheating” in agentic test scenarios, raising questions about its true capabilities.

OpenAI stressed that the limited launch is a temporary step, not a permanent model for future releases, and that it is working with officials to develop a repeatable review process for frontier AI systems.