OpenAI to widen access to GPT-5.6 after US-requested pause

OpenAI said it will make its GPT-5.6 Sol, Terra and Luna models publicly available on Thursday, ending an initial restriction that kept access to a limited set of partners for about two weeks. The release affects developers, enterprises and security users seeking access to the company’s latest artificial intelligence systems, including GPT-5.6 Sol, which OpenAI has described as its most capable model to date.

The company announced the GPT-5.6 family in June, but said at the time that it had agreed to begin with a narrower rollout at the request of the U.S. government. OpenAI said participation in that early access group had been shared with government officials, and described the initial users as trusted partners.

Sam Altman, OpenAI’s chief executive, signaled the wider availability late Tuesday in a post on X, writing, “Happy building.” OpenAI also said in a separate X post that it was expanding preview access to the GPT-5.6 models worldwide.

Government review shaped the initial rollout

OpenAI said in June that it supported broader distribution of its models and intended to increase availability in the following weeks. The company also cautioned that a government access process should not become the standing model for releases, saying at the time that such a system could keep advanced tools away from users, developers, enterprises, cyber defenders and international partners that need them.

The restriction reflected a broader shift in U.S. policy toward advanced AI releases. President Donald Trump signed an executive order in June asking AI model developers to voluntarily provide frontier systems to the government so officials can assess their capabilities before a full public launch. The order gave federal agencies 60 days to create an evaluation process.

Under that approach, a developer can give the government access to a model before general release, allowing officials to examine its capabilities and risks. OpenAI said in June that it was working with the government on a framework for such reviews and on a repeatable process for future model launches.

Anthropic dispute preceded the OpenAI release

OpenAI’s wider release follows a separate clash between Anthropic and the U.S. government over access to Anthropic’s Claude Fable 5 and Mythos 5 models. Anthropic had disabled access to those systems to comply with an export-control directive, according to CNBC. The U.S. Department of Commerce lifted that directive late last month, after which Anthropic restored access.

The OpenAI release adds another data point in the evolving relationship between AI developers and U.S. authorities as the government seeks more visibility into cutting-edge models before they reach broad use. For companies building on top of these systems, the timing and scope of releases can affect product development, cybersecurity work and enterprise adoption schedules.

OpenAI has said GPT-5.6 Sol is stronger than its prior models and has improved capabilities in coding, biology and cybersecurity. The company has not provided additional quantitative performance figures in the announcement referenced Tuesday.

This story draws on original reporting from CNBC.