OpenAI Unveils GPT-5.6 Series in Restricted Launch

OpenAI has released its GPT-5.6 models—Sol, Terra, and Luna—but only to government-approved customers, sparking debate over AI governance.

OpenAI has introduced its latest generation of models, the GPT-5.6 series, comprising three distinct variants: Sol, Terra, and Luna. Sol, the flagship model, represents the company's most advanced AI to date, boasting improvements in coding, biology, and cybersecurity through enhanced agentic capabilities. Terra offers performance comparable to GPT-5.5 but at half the cost, while Luna provides strong capability at the lowest price point yet. A new 'max' reasoning effort and an 'ultra' mode, which employs sub-agents for complex tasks, further distinguish the lineup.

Despite these advancements, the rollout is anything but straightforward. The Trump administration has intervened, limiting the release to a select group of government-approved customers and trusted partners. OpenAI, while complying, has expressed reservations, arguing that such government access processes should not become the long-term default. The company contends that restricting access keeps powerful tools from developers, enterprises, and cyber defenders who need them, and views this short-term measure as a path to broader availability in the coming weeks.

The administration's move follows a June 2 executive order establishing a framework for benchmarking and assessing new AI models before public release. This is not the first instance of government intervention; Anthropic was previously forced to withdraw its Claude Fable 5 and Mythos 5 models despite adhering to voluntary review processes. OpenAI asserts that GPT-5.6 Sol incorporates its most robust safety stack to date, with protections against high-risk activity and misuse, and has been hardened against real-world attacks. The model is designed to assist in identifying and fixing vulnerabilities rather than enabling end-to-end attacks, balancing defensive benefits with prohibitions on offensive use.