On June 26, 2026, OpenAI announced a preview of GPT-5.6, a next-generation model family in three sizes: Sol, the flagship for the hardest problems; Terra, a balanced model for high-volume work; and Luna, a fast, low-cost model for everyday tasks. OpenAI calls GPT-5.6 Sol its strongest model yet, highlighting improved agentic capabilities in coding, biology, and cybersecurity.
The launch is unusual for its rollout. As part of ongoing engagement with the US government, OpenAI previewed the models’ capabilities to the administration ahead of launch and, at the government’s request, started with a limited preview for a small group of trusted partners whose participation was shared with the government. The restricted rollout follows the June 2, 2026 executive order on advanced AI innovation and security, which directed federal agencies to build a process for assessing the capabilities of new frontier models before public release - the first time the US government preemptively asked an American AI company to limit a model launch, coming two weeks after Anthropic suspended access to two of its models under an export control directive.
Under OpenAI’s Preparedness Framework, the GPT-5.6 Preview system card classifies all three models High in both Cybersecurity and Biological/Chemical capability, below the Critical threshold, and describes a layered safety stack including safety training, new activation classifiers for sensitive domains, real-time output blocking, and automated monitoring. Pricing per million tokens is 5 dollars input and 30 dollars output for Sol, 2.50 and 15 for Terra, and 1 and 6 for Luna, alongside more predictable prompt caching with explicit cache breakpoints and a 30-minute minimum cache life.
OpenAI said the preview is a short-term step toward general availability in the coming weeks, and publicly pushed back on the arrangement, saying it does not believe this kind of government access process should become the long-term default. The release marks a visible shift in how frontier models reach the market as AI capability and national security policy converge.