OpenAI will publicly release all three GPT-5.6 variants on Thursday, July 9, after keeping the model family restricted to a select group of partners for two weeks. The announcement came via OpenAI's X account, confirming reporting from Axios that the U.S. Department of Commerce had approved the broader rollout.

The series comprises three models: Sol, Terra, and Luna, first shown to developers on June 26. Sol is OpenAI's flagship, designed for developers and enterprises, and promises state-of-the-art performance in reasoning, coding, and scientific tasks, with strengthened cybersecurity capability. It costs $5 per million input tokens and $30 per million output tokens — OpenAI's highest price point. Terra targets general-purpose use, delivering performance comparable to GPT-5.5 at roughly half the cost: $2.50 per million input tokens and $15 per million output tokens. Luna is the budget option, at $1 per million input tokens and $6 per million output tokens.

The two-week delay between preview and public launch stems from a Trump administration request. In early June, the president signed an AI cybersecurity order asking companies to voluntarily submit their most advanced models to government review 30 days before public release. OpenAI complied by limiting GPT-5.6's late-June debut to trusted partners instead of the general public, while publishing a system card on its Deployment Safety Hub and a help center article. At the time, OpenAI stated publicly that such restrictions "shouldn't be the norm."

The review itself ran through the Department of Commerce's Center for AI Standards and Innovation, which conducted additional testing on GPT-5.6. OpenAI sent technical staff to Washington to address the agency's questions directly. That process, combined with further meetings between the company and administration officials, produced the clearance enabling Thursday's wider release.

OpenAI is not alone in encountering this review mechanism. Anthropic previously blocked access to its Mythos cybersecurity model and Fable model after being ordered to restrict foreign national users. The company later regained permission to redeploy Mythos 5, and eventually Fable 5, once government concerns were resolved.

The pattern across both companies shows a consistent sequence: a frontier model is built, shown quietly to a narrow group of partners or held back from certain user classes, subjected to federal review, then cleared for wider release once officials sign off. This represents a material shift from the voluntary, industry-led disclosure norms that governed earlier model launches, where companies typically set their own release timeline and consulted government mainly on request rather than under a standing 30-day review requirement.

The deeper question — whether this becomes a permanent feature of the U.S. regulatory landscape or a temporary arrangement tied to this particular administration's cybersecurity order — remains open. OpenAI's own language, that such restrictions "shouldn't be the norm," suggests the company views the current arrangement as an exception rather than a model it would support long-term. That positioning was published while the company was still complying with the order, which indicates the friction is being managed rather than fundamentally resolved.

For enterprise teams, the more immediate consideration is pricing strategy. Terra's positioning — GPT-5.5-class performance at half the cost — follows a pattern visible across recent model generations: each new release pushes down the price of "good enough" capability while a smaller, pricier flagship absorbs the frontier performance gains. Sol's $30-per-million-output-token rate sits well above Terra and Luna, presumably reserved for workloads where the cybersecurity and reasoning improvements justify the premium. Organizations running cost-sensitive inference at scale will likely default to Luna or Terra, reserving Sol for the narrower set of tasks where its stated strengths genuinely move the needle.

The more consequential thread here, though, is not the pricing tiers but the review mechanism itself. A 30-day government preview window for frontier models, now applied twice across two major labs within weeks, resembles the kind of regulatory infrastructure that tends to persist once established. Whether that holds true will depend less on this week's clearance than on how the next model cycle handles the same request.