OpenAI Halts GPT-5.6 After Government Pressure

OpenAI Halts GPT-5.6 Release After Government Pressure

OpenAI is limiting the release of its new AI models to a "small group of trusted partners" at the request of the U.S. government, the company announced on Friday.

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

The administration's request comes as the U.S. government exerts new pressure on AI companies to limit their most advanced systems. After Anthropic released its most powerful public model, Fable 5, the administration ordered the company to revoke access for any foreign nationals, leading Anthropic to completely withdraw the model.

This incident raises questions about the extent of government power over AI model releases. Dean Ball, a former AI advisor at the White House and future employee of OpenAI, claims that the recent executive order from President Trump—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 involuntary licensing regime for cutting-edge AI, resulting in severe restrictions.

The issue becomes more complicated, Ball argues, when the government lacks clearly defined safety standards, which could lead to endless launch delays that not only give China an advantage in the AI race but also jeopardize the billions of dollars invested in AI infrastructure.

Although OpenAI complied with the administration's request this time, the AI company has made it clear that it is not satisfied with the arrangement.

"We do not believe that this type of government access process should become the long-term default," reads a blog post published on Friday. "It prevents the best tools from reaching users, developers, businesses, cybersecurity advocates, and global partners who need them."

OpenAI described the preview as a "short-term measure" that will put GPT-5.6 on the path to broader availability in the coming weeks, as the company works with the administration to develop a new executive order framework on cybersecurity, as well as a "repeatable process for future model releases."

GPT-5.6 Sol Specifications

OpenAI claims that GPT-5.6 Sol is its most powerful model to date, with enhanced agentic capabilities in programming, biology, and cybersecurity. Sol introduces a "max" reasoning effort mode and an "ultra" mode that uses coordinated sub-agents to tackle very complex tasks (which significantly increases your token usage).

According to OpenAI, GPT-5.6 excels in several benchmarks, including being slightly better in programming workflows than Anthropic's Claude Mythos 5, which has also been effectively banned this month by the Trump administration. OpenAI asserts that GPT-5.6 Sol is also competitive with the Mythos preview but uses one-third of the output tokens.

To alleviate concerns regarding the safety of its powerful models, OpenAI states that Sol includes its most robust safety stack to date. According to OpenAI, it is heavily fortified against adversarial attacks and intentionally optimized to favor defensive cybersecurity work over offensive exploits. In other words, it is designed to be difficult to circumvent while prioritizing teaching users about defense against exploits rather than how to hack systems.

OpenAI also clarifies that its safety guardrails are integrated directly into the core model's behavior rather than relying on a separate filter. The company is likely trying to avoid the trap that ensnared Anthropic with Fable 5. During the brief moments when Fable 5 was available, whenever the model's classifiers detected a high-risk topic—such as cybersecurity, biology, or chemistry—it did not merely block the request; it redirected the query to an older model. This excessively cautious flow and invisible redirection led to numerous false positives and negative user feedback.

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

GPT-5.6 comes in three sizes with tiered pricing: Sol costs $5 per million input tokens and $30 per million output tokens; Terra costs half of that; and Luna costs $1 and $6, respectively. OpenAI also indicates that it has improved the request cache to make repeated queries less costly and more predictable.