OpenAI to Publicly Release GPT-5.6 Models After U.S. Government Clears the Way

OpenAI will release its GPT-5.6 Sol, Terra, and Luna models to the public on Thursday, ending a roughly two-week period during which the company had restricted access to a limited set of trusted partners at the request of the U.S. government.

The U.S. Department of Commerce cleared OpenAI to proceed with a broad rollout, according to a source familiar with the matter, following additional testing and meetings between company representatives and government officials. OpenAI, the White House, and the Department of Commerce did not immediately respond to requests for comment on Tuesday.

"Happy building," CEO Sam Altman wrote in a post on X late Tuesday, signaling the expansion of preview access globally.

OpenAI first announced the three models in June, at which point the company agreed to limit distribution to organizations whose "participation has been shared with the government." The company did not disclose the names of those partner organizations.

In a blog post accompanying the June announcement, OpenAI said it "believes in broad access" and expressed intent to make the models more widely available over the coming weeks. The company added that it did not view the restricted rollout as a desirable long-term model. "We don't believe this kind of government access process should become the long-term default," OpenAI said at the time. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them."

OpenAI described GPT-5.6 Sol as its "strongest model yet," with enhanced capabilities across coding, biology, and cybersecurity, according to a company blog post.

The public release comes amid a broader shift in how the Trump administration is engaging with frontier AI development. President Donald Trump signed an AI executive order in June directing model developers to voluntarily provide advanced models to the government for capability assessments before full public release. The order also gave federal agencies 60 days to build out an evaluation process. OpenAI said in June it is working alongside the government to establish a repeatable framework for future model releases.

OpenAI's chief rival, Anthropic, navigated a similar restriction period. The company was required to suspend access to its Claude Fable 5 and Mythos 5 models to comply with an export control directive issued by the U.S. Department of Commerce. That directive was lifted late last month, restoring access after a weeks-long gap that affected users worldwide.

The sequential nature of these restrictions — first Anthropic, then OpenAI — signals an emerging pattern of government-mandated review windows for frontier AI models, a dynamic that analysts say is creating openings for overseas competitors. Chinese AI firm Zhipu, trading as Knowledge Atlas Technology JSC, launched its GLM 5.2 model last month, offering it as a free download that enterprises can fine-tune and run on their own servers.

How consistently the government applies this review process to future model releases, and whether the 60-day agency evaluation timeline produces a durable assessment framework, will determine whether Tuesday's resolution represents a workable template or a one-off accommodation.