OpenAI previewed its GPT-5.6 family on Friday, gating access behind a list of partners cleared by the Trump administration in what’s the first flagship U.S. frontier model to launch under a Washington-managed allowlist. The company calls the arrangement a “short-term step” taken at the administration’s request, with broader general availability expected within weeks.

The lineup is tiered. Sol is the flagship, which OpenAI says sets a new state of the art on the Terminal-Bench 2.1 coding benchmark and posts gains in biology and cybersecurity workflows. Terra sits in the middle. Luna is the cheapest and fastest. Pricing, per TechCrunch: Sol runs $5 per million input tokens and $30 per million output; Terra is half that, at $2.50 and $15; Luna lands at $1 and $6. The GPT-5.6 Preview System Card states the models don’t reach OpenAI’s “critical” cybersecurity threshold, and that Sol and Terra couldn’t run autonomous end-to-end attacks against hardened targets in testing.

Bloomberg first reported that the partner names had been approved by the U.S. government. CNBC noted that Anthropic recently disabled two of its latest models to comply with an export-control directive from the same administration. Two frontier labs, two access regimes shaped by the same hand.

OpenAI says it’s collaborating with the administration on a new executive-order framework for cybersecurity and a “repeatable process for future model releases.” Dean Ball, the former White House AI adviser and incoming OpenAI employee, told TechCrunch that a recent Trump executive order asking companies to voluntarily submit advanced models for review up to 30 days before release has produced a de facto licensing regime.

A voluntary review, with cleared partners and a repeatable process, is the architecture of licensing under another name.

Sources