Commerce Department Clears OpenAI's GPT-5.6 for Broad Public Launch on July 9
- The Commerce Department's Center for AI Standards and Innovation cleared GPT-5.6 for broad public release after weeks of government testing [1]
- GPT-5.6 had been restricted to roughly 20 government-vetted partners since June 26, the first time a U.S. AI lab gated a frontier model behind a state-approved roster [2]
- Three model tiers launch Thursday: Sol (flagship, $5/$30 per million tokens), Terra ($2.50/$15), and Luna ($1/$6) [3]
- OpenAI said it does not believe government-gated access 'should become the long-term default' for frontier AI releases [4]
- The approval follows Trump's June AI cybersecurity order requiring labs to submit powerful models to the government 30 days before public release [5]
The U.S. Department of Commerce has given OpenAI the green light for a broad public launch of its GPT-5.6 model family, ending a first-of-its-kind restricted preview that limited access to roughly 20 government-approved organizations for nearly two weeks. All three model tiers — the flagship Sol, the mid-range Terra, and the budget Luna — will become publicly available on Thursday, July 9 [1].
The clearance was granted after testing conducted by the Commerce Department's Center for AI Standards and Innovation, with OpenAI dispatching technical experts to Washington to address agency questions about the models' capabilities in sensitive domains including cybersecurity and biology [5]. The review marks a precedent-setting moment for the American AI industry: GPT-5.6 is the first frontier model from a leading U.S. lab to be released on the government's timeline rather than the company's own [2].
OpenAI first unveiled the GPT-5.6 series on June 26 but was forced to limit initial access at the Trump administration's request. The government approved partners on a customer-by-customer basis, with agencies including the Office of the National Cyber Director and the Office of Science and Technology Policy involved in the review [2]. The company's own system card rates all three models as 'High capability' for cybersecurity and biological/chemical risks while remaining below critical thresholds for AI self-improvement [6].
The Models
GPT-5.6 ships in three tiers designed for different price and performance points. Sol, the flagship, is built for complex reasoning and agentic workloads and includes a 'max reasoning effort' mode that gives the model extended processing time for difficult problems. OpenAI said Sol is competitive with Anthropic's Mythos Preview on the ExploitBench cybersecurity benchmark [5].
Terra occupies the mid-range, offering performance competitive with the prior-generation GPT-5.5 at roughly half the cost. Luna is the smallest and fastest model in the series, optimized for high-volume, cost-sensitive applications [3].
API pricing reflects the tiered approach: Sol costs $5 per million input tokens and $30 per million output tokens; Terra runs $2.50/$15; Luna is $1/$6. A Sol 'Fast' mode capable of up to 750 tokens per second is available at $12.50/$75 [3].
The Government Review
The restricted preview originated from a Trump administration executive order issued in June that established a voluntary framework for AI developers to submit 'covered frontier models' to the government for up to 30 days before public release [5]. OpenAI became the first major lab to go through the process.
During the review period, access was limited to a short list of organizations whose identities OpenAI shared with the government — approximately 20 partners vetted on a case-by-case basis. The specific names were not publicly disclosed [2]. The Commerce Department's Center for AI Standards and Innovation led the technical evaluation, focusing on the models' capabilities in cybersecurity, biological and chemical domains, and potential for AI self-improvement [6].
The review process has also touched other frontier labs. Anthropic's Mythos 5 and Fable 5 models were temporarily disabled following a June 12 export control order, with restrictions lifted only after the company implemented additional safeguards [5].
Why It Matters
The GPT-5.6 clearance establishes a live template for how Washington intends to oversee the release of increasingly powerful AI systems. Rather than legislating blanket rules, the Trump administration is negotiating access on a model-by-model, company-by-company basis — a regulatory posture that gives the executive branch significant discretion over which technologies reach the public and when [1].
For OpenAI, the approval removes a competitive disadvantage. During the restricted preview, rivals including Anthropic and xAI's Grok continued to iterate on their own models. Elon Musk's xAI announced that Grok 4.5 would become publicly available at the same time as GPT-5.6's broad rollout [5].
The episode also signals a new compliance burden for AI developers. Teams building on frontier models now face requirements including safety evidence documentation, red-team reports, approved-user lists, and customer eligibility verification — governance overhead that goes well beyond traditional API migration workflows [6].
What's Next
OpenAI said it is 'expanding preview access globally now,' with full general availability across ChatGPT and the API beginning Thursday [3]. The company has not disclosed whether future model releases will go through the same government review process, though the June executive order's 30-day submission framework remains in effect.
The broader question is whether this case-by-case approval model scales. With multiple U.S. labs — and Chinese competitors — racing to release increasingly capable systems, the Commerce Department's Center for AI Standards and Innovation faces a growing pipeline of models that will require evaluation. China has reportedly imposed its own restrictions on overseas access to advanced AI models, mirroring Washington's approach from the opposite direction [5].
Further sources
The stories that matter, in one email. Free — unsubscribe anytime.