OpenAI has begun previewing its GPT-5.6 model series to a select group of trusted partners, with broader availability expected in the coming weeks. The lineup includes three variants: Sol as the flagship model, Terra optimized for everyday use, and Luna as the budget-friendly option. OpenAI emphasized that Terra delivers GPT-5.5-level performance at roughly half the cost.

The preview comes after the US government received early access to GPT-5.6, following President Trump's recent AI cybersecurity executive order requiring companies to submit their most powerful models for voluntary review 30 days before public release. OpenAI stressed this government-access approach shouldn't become a long-term default, but serves as a necessary short-term step to ensure public deployment.

Sol features enhanced cybersecurity capabilities and a new "max" reasoning effort that allows deeper analysis. The company invested 700,000 GPU hours identifying universal jailbreak vulnerabilities and implementing safeguards across all variants. Sol's pricing at $5 per million input tokens and $30 per million output tokens undercuts Anthropic's previous Fable model costs.

This rollout follows Anthropic's recent suspension of access to its Mythos 5 and Fable 5 models after government concerns about potential jailbreaking. Meta remains the sole major AI developer refusing early government model evaluations, highlighting ongoing industry tensions around regulatory compliance and competitive positioning.