GPT-5.6 Sol, Terra, and Luna: OpenAI's Most Powerful Model Just Dropped, but the White House Picks Who Gets It

For the first time ever, the U.S. government has preemptively asked an American AI company to restrict the launch of a model before it reaches the public. OpenAI complied on Friday, June 26, by unveiling GPT-5.6 in a limited preview reserved for roughly 20 government-approved organizations.

GPT-5.6 Sol, Terra, and Luna: three tiers, one family

The GPT-5.6 lineup splits into three variants. Sol is the flagship for complex reasoning, coding, and cybersecurity. Terra matches GPT-5.5 performance at half the cost. Luna delivers speed and affordability for high-volume tasks.

Pricing runs at $5/$30 per million tokens (input/output) for Sol, $2.50/$15 for Terra, and $1/$6 for Luna. Two new reasoning modes, « max » and « ultra, » are also rolling out. Ultra delegates work across multiple sub-agents for harder problems.

Why Washington put the brakes on GPT-5.6

The White House cited national security concerns. According to Axios reporting, the administration considers GPT-5.6 to have « Mythos-like » capability. That same Anthropic Mythos triggered the forced shutdown of Claude Fable 5 on June 12, when the Commerce Department pulled the plug for all users worldwide.

OpenAI pushed back publicly. The company stated that this kind of government gatekeeping « keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. » Sam Altman expects broader access within weeks.

GPT-5.6 Sol benchmarks push the frontier

On TerminalBench 2.1, Sol in ultra mode scored 91.91%, up from GPT-5.5's 83.4% and ahead of Claude Mythos 5's 88%. The system card rates all three models as « High » capability in cybersecurity and biological risk, without crossing the « Critical » threshold.

But the bigger story isn't benchmarks. Washington now treats frontier AI models the way it treats sensitive defense technology. For developers outside the U.S. and enterprises that rely on cutting-edge tools, this precedent could delay access by weeks or longer.

A new normal for AI launches

In the span of two weeks, the government recalled Anthropic's Fable 5 and forced OpenAI into a gated rollout. Recent decisions suggest that unrestricted frontier-model launches may soon come to an end. A stable regulatory framework, however, has yet to replace the current case-by-case approach.