Enterprise

OpenAI gates GPT-5.6 launch to U.S.-approved partners at the Trump administration's request

The Sol, Terra, and Luna models go to roughly the same partner pool Anthropic was forced into two weeks earlier — a new pre-release vetting layer between frontier labs and their enterprise customers.

By Asher Glick · Enterprise software desk · 2026-06-29 · 4 min read

Photo: Unsplash / Andy Feliciotti — The United States Capitol dome in Washington, D.C.

OpenAI on Friday previewed three new models, GPT-5.6 Sol, Terra, and Luna, and confirmed the rollout will be restricted to a list of trusted partners approved by the U.S. government at the Trump administration’s request. The same day, Anthropic was finally cleared to release Mythos 5 to roughly 100 companies and federal agencies, ending a standoff that had suspended both Fable 5 and Mythos 5 for about two weeks. Two of the three frontier labs in the United States are now shipping their latest models through a Washington-approved partner pool.

Sol is the flagship, priced at $5 input and $30 output per million tokens. Terra slots in at half that, $2.50 input and $15 output. Luna, the fastest and most cost-efficient of the three, lands at $1 and $6. General availability is expected in the coming weeks.

The trigger, per OpenAI’s deployment-safety note, is the Preparedness Framework: Sol, Terra, and Luna all rate High in both Cybersecurity and Biological and Chemical risk. On ExploitBench, OpenAI says Sol is competitive with Mythos Preview using only about a third of the output tokens, and characterizes the model as better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks. That’s the technical posture the company is bringing to its government reviewers.

The political posture is more interesting. OpenAI states plainly that it doesn’t believe this kind of government access process should become the long-term default, and says it’s working with the administration on a cyber executive order framework and a repeatable process for future releases. Dean Ball, the former White House AI adviser who’s set to join OpenAI, has argued that Trump’s executive order, which asks certain companies to voluntarily submit advanced models for review up to 30 days before release, has produced a de facto involuntary licensing regime, per TechCrunch’s Rebecca Bellan.

That’s the structural shift worth naming. Frontier procurement now runs through Washington before it reaches the buyer. The voluntary label is doing a lot of work.

Sources