Artificial intelligence (AI) giant OpenAI announced on June 26 the preview release of its new class of models, which, for the time being, will be limited to a small group of users approved by the Trump administration.
The ChatGPT-maker is launching a limited preview of Sol, Terra, and Luna under the banner of GPT-5.6. OpenAI explained that Sol is the firm’s new flagship model, while Terra is balanced for “everyday work” and Luna is its “fast and affordable model.”
For now, the models will be available only to a small group of partners approved by the executive branch after a request from the Trump administration following conversations with government officials, OpenAI said. The company will continue testing the models during the preview while pushing for a broader release.
“We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them,” OpenAI said in a press release.
“We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks, while we work with the Administration to develop the cyber Executive Order framework and a repeatable process for future model releases.”
The announcement comes just weeks after competitor Anthropic initially released its new flagship Claude Fable 5 and Claude Mythos 5 models. In April, Anthropic cooperated with the Trump administration in a preview release of Mythos to a select group of cybersecurity testers and critical infrastructure partners.
Then on June 12, the Trump administration ordered Anthropic to ban all foreign nationals—both inside and outside the United States—from accessing the two models. Anthropic then banned access to all users.
Both models are unavailable for the time being.
President Donald Trump said he found the action responsible, and claimed one of Anthropic’s competitors, which he didn’t name, flagged Fable 5 to his team.
In a statement at the time, Anthropic said, “Our understanding is that the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5.”
During the initial limited rollout of Mythos in April, Anthropic said the model was capable of finding critical software exploits that, in some cases, are unknown to developers at launch, referred to as “zero-day” exploits.
While OpenAI said its ChatGPT-5.6 Sol is its most advanced model yet, Sol only found the “building blocks of an exploit” when tested with major internet web browsers.
“Still, benchmark thresholds cannot capture every way a model may be used or combined with other tools,” OpenAI said.
“That uncertainty, along with the model’s broader step change in capabilities, is why we are pairing the model’s increased capabilities with stronger safeguards and a phased release.”
