OpenAI moved closer to a public GPT-5.6 launch after federal officials cleared broader access to the new model family. The decision ends weeks of limited availability and signals a less restrictive path for frontier AI releases in the market. The launch now places model safety, federal review, and global competition at the center of the fast-moving AI industry.
U.S. approval clears a wider rollout
OpenAI expects to release GPT-5.6 publicly on Thursday, after additional testing and direct discussions with federal officials in Washington. The approval followed Commerce Department review of the model family and its reported high-risk technical capabilities before wider availability. Axios reported the clearance, citing a source familiar with discussions between the company and government officials this week.
The company had restricted early access to roughly 20 partners vetted through a federal process before the clearance took effect. That approach followed a government request for staged availability while safety and security reviews continued in Washington this month. OpenAI did not identify the partner list, and officials did not immediately issue detailed public comments on the decision.
The OpenAI release covers three models, named Sol, Terra, and Luna, according to public company materials and preview details. Sol serves as the top model, while Terra targets lower-cost enterprise uses and business workloads for cost-sensitive customers at scale.
Luna offers the fastest and cheapest option for high-volume tasks, customer support, and routine deployments across teams and services.
Federal scrutiny focused on frontier risks
United States officials examined GPT-5.6 because advanced systems can raise cyber, biological, and security concerns before launch approval. The review focused on capabilities that could matter for national security, public safety, and technical misuse by outside actors. The process also tested how agencies can assess frontier models before broad commercial access begins for public users.
President Donald Trump signed a June 2 order establishing voluntary pre-release checks for powerful AI systems before broad release. However, the GPT-5.6 process went beyond that framework because access stayed under federal management during the review for weeks.
That created a temporary access list while officials completed evaluations and held technical meetings with OpenAI staff in Washington.
OpenAI has said it supports broad access while continuing work on safety controls and deployment safeguards for commercial users. The company said it “believes in broad access” when it announced the model family last month to users. However, the temporary limits showed how federal oversight can shape OpenAI product availability before public launch in the United States.
Rival models and global pressure shape the decision
The decision followed similar restrictions affecting Anthropic models during the same regulatory period in Washington and other agencies. Commerce action had limited access to Claude Fable 5 and Mythos 5 over security concerns and export-control issues last month. Fable 5 later returned to availability, while Mythos 5 remained limited to selected American organizations under federal terms.
The pause also arrived as Chinese developers promoted more accessible models for enterprise and research users outside the United States. Zhipu, operating as Knowledge Atlas Technology JSC, released GLM 5.2 last month for external use and testing worldwide. The company made the model free to download, fine-tune, and run on private servers for internal enterprise deployment and research.
OpenAI now enters the market with a clearer path, but scrutiny of frontier AI remains active after clearance. The GPT-5.6 rollout will test how OpenAI manages safety obligations while serving global customers at scale across markets. It will also show whether federal review becomes a regular step for advanced model releases in the United States.
