OpenAI limits GPT-5.6 rollout after government request, says restrictions shouldn’t be the norm

OpenAI is limiting the release of its latest AI models to a “small group of trusted partners” at the behest of the U.S. government, the company said Friday.

The next generation GPT-5.6 series includes Sol, the flagship model; Terra, a more balanced model for everyday use; and Luna, a faster, cheaper option. Although Sol is the company’s most powerful model, the Trump administration has restricted the release of all three models. OpenAI said the preview is limited to partners “whose participation has been shared with the government.”

The administration’s request comes as the U.S. government puts new pressure on AI companies to limit their most advanced systems. After Anthropic released its most powerful public model Fable 5, the government ordered the company to remove access for any foreigner, prompting Anthropic to scrap the model entirely.

The incident has raised questions about how much power the government should have over the release of AI models. Dean Ball, a former White House AI adviser and soon-to-be OpenAI contributor, says President Trump’s recent executive order — asking certain AI companies to voluntarily submit their most advanced models to government review up to 30 days before release — has created a de facto involuntary licensing regime for borderline AI, leading to heavy-handed restrictions.

The problem is compounded, Ball argues, when the government doesn’t have clearly defined safety standards, which could lead to endless launch delays, which would not only give China a hand in the AI ​​race but also jeopardize the billions of dollars going toward building AI infrastructure.

And although OpenAI did what the administration asked this time, the AI ​​company made it clear that it was not happy with the arrangement.

“We do not believe that these types of government access procedures should become the standard in the long term,” a statement said Friday blog post. “It keeps the best tools for users, developers, enterprises, cyber defenders and global partners who need them.”

OpenAI called the preview a “short-term step” that will put GPT-5.6 on the path to broader availability in the coming weeks, as the company works with the government to develop a new framework for cybersecurity, as well as a “repeatable process for future model releases.”

GPT-5.6 Sol Specifications

OpenAI says GPT-5.6 Sol is the strongest model yet, with enhanced agentic capabilities in encryption, biology and cybersecurity. Sol introduces a “max” reasoning effort mode and an “ultra” mode that uses coordinated sub-agents to solve highly complex tasks (exactly the kind of neat trick that makes your token usage skyrocket).

GPT-5.6 excels in several benchmarks, OpenAI says, including being slightly better at coding workflows than Anthropic’s Claude Mythos 5, which the Trump administration also effectively banned this month. OpenAI says GPT-5.6 Sol is also competitive with the Mythos preview, but uses a third of the output tokens.

To allay fears that its powerful models are insecure, OpenAI says Sol includes the most robust security stack yet. It is, OpenAI says, heavily hardened against adversarial attacks and deliberately optimized to favor defensive cybersecurity work over offensive exploits. In other words, it is designed to be difficult to jailbreak while prioritizing how users can defend against exploits, rather than how to hack into systems.

OpenAI also says that the safety bars are built directly into the behavior of the core model, rather than relying on a separate filter on top. The company is likely trying to avoid the pitfall Anthropic fell into with Fable 5. In the brief moments that Fable 5 was available, it not only blocked the prompt if the model’s classifiers detected a high-risk topic such as cybersecurity, biology, or chemistry; it would route the request to an older model. The very careful flow and invisible downrouting led to many false positives and user responses.

While the GPT-5.6 models will initially only be available to a select group of partners, OpenAI plans to soon make them more widely available to people using ChatGPT, Codex, and the API.

GPT-5.6 is available in three sizes with tiered pricing: Sol costs $5 per million input tokens and $30 per million output tokens; Terra costs half that; and Luna cost $1 and $6 respectively. OpenAI says it has also improved prompt caching to make repeating prompts cheaper and more predictable.