OpenAI limits GPT-5.6 release after government request, says restrictions should not be common

OpenAI is limiting the release of its new AI models to a “small group of trusted partners” at the urging of the US government, the company said on Friday.

The next generation GPT-5.6 lineup includes the Sol, its flagship model; Terra, a balanced model for everyday use; and Luna, a faster, lower-cost option. Although Sol is the company’s most powerful model, the Trump administration has restricted the rollout of all three. OpenAI said the preview was limited to partners “whose collaborations are shared with governments.”

The administration’s request comes as the US government puts new pressure on AI companies to curb their most advanced systems. After Anthropic released its most powerful public model Fable 5, management ordered the company to remove access to any outsiders, prompting Anthropic to drop the model entirely.

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

The problem is compounded, Ball says, when the government does not have clearly defined security standards, which may lead to endless delays in the launch that may not only help China in the AI ​​race, but also jeopardize the billions of dollars going to the construction of AI infrastructure.

And while OpenAI did as the administration asked for this, the AI ​​company made it clear that it was not happy with the plan.

“We do not believe that this type of government outreach should be a long-term mistake,” the Friday blog post read. “It houses the best tools from users, developers, businesses, Internet defenders, and global partners that they need.”

OpenAI called the preview an “interim step” that will put GPT-5.6 on the path to wider availability in the coming weeks, as the company works with management to develop a new high-order framework for cybersecurity, and “an iterative process for future model releases.”

Details of GPT-5.6 Sol

OpenAI says GPT-5.6 Sol is its most powerful model yet, with advanced agent capabilities in coding, biology, and cybersecurity. Sol introduces a “big” thinking mode and an “ultra” mode that uses integrated subagents to solve more complex tasks (just a neat trick that sends your token usage up).

GPT-5.6 excels in several benchmarks, OpenAI said, including being slightly better at documenting workflows than Anthropic’s Claude Mythos 5, which the Trump administration effectively blocked 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 any fears that its powerful models are unsafe, OpenAI says Sol includes its strongest security stack yet. OpenAI says, it is more robust against adversary attacks and is purpose-built to favor defensive cybersecurity operations over offensive operations. In other words, it’s designed to make it harder to jailbreak, while prioritizing showing users how to protect against exploits, rather than hacking systems.

OpenAI also claims that its security guardrails are built directly into the behavior of the core model, rather than relying on a separate filter on top of it. The firm is probably trying to avoid the trap that caught Anthropic with Fable 5. In the short times when Fable 5 was available, whenever the model designers found a very dangerous topic – such as cybersecurity, biology, or chemistry – it would not just block the information; will move the application to the old model. All the flow of over-monitoring and invisible routing has led to a lot of false positives and user backlash.

Although GPT-5.6 models are initially only available to a select group of partners, OpenAI plans to make them more widely available to users of ChatGPT, Codex, and the API soon.

GPT-5.6 comes in three price sizes: Sol costs $5 for a million input tokens and $30 for a million output tokens; Terra costs half that; and Luna costs $1 and $6, respectively. OpenAI says it has also improved fast caching to make repetitive data cheaper and more predictable.