Summary created by Smart Answers AI
In summary:
- OpenAI has released its most powerful GPT-5.6 models (Sol, Terra, and Luna) but only to trusted partners and organizations following a U.S. government request for limited rollout.
- PCWorld reports that the flagship Sol model directly competes with Anthropic’s restricted Fable and Mythos models, featuring enhanced layered safeguards against misuse.
- A wider public release remains uncertain due to regulatory concerns about the models’ advanced capabilities and potential security implications.
The U.S. government ban on Anthropic’s powerful Fable and Mythos models has everybody spooked. That’s why it’s a little surprising that OpenAI, Anthropic’s biggest rival, is slowly rolling out its latest and most powerful GPT models.
Yes, the GPT-5.6 series of models is here, but not for everyone. It comes in three flavors: the flagship Sol model, the everyday Terra, and the fast-and-efficient Luna. Those models map roughly to the Opus, Sonnet, and Haiku configuration of Anthropic’s Claude models.
That said, Sol is really more of a competitor to Fable and Mythos, the two Claude models that were slapped with expert restrictions by the Trump Administration earlier this month over fears they were too dangerous for broad release.
Like Fable and particularly Mythos (which was restricted to only a small group of handpicked users and organizations even before the U.S. ban), GPT-5.6 Sol is being billed as OpenAI’s “strongest model yet,” complete with robust coding, biology, and cybersecurity capabilities.
But while Anthropic put Claude Fable into general release before being ordered to pull it down days later, OpenAI said its GPT-5.6 models will start off in a “limited preview” for just a “select group of trusted partners and organizations” – meaning, basically, not you.
In a press release announcing the new models, OpenAI said it had notified the U.S. government of its “plans and the models’ capabilities,” and was limiting the rollout of the models “at their request.”
While allowing that the “limited” release for GPT-5.6 will help ensure “the strongest path to broader availability in the coming weeks,” OpenAI warned that “this kind of government access process … keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”
Anticipating the concerns raised by Anthropic’s Fable and Mythos models, which are said to have breached sensitive national security systems “in hours,” OpenAI went to great lengths to detail GPT 5.6’s “layered” safeguards, including against jailbreaking and “real-world adversarial pressure.”
OpenAI said it anticipates that GPT-5.6 Sol, Terra, and Luna will be released “soon” to general users. But just as Claude Fable and Mythos remain in a regulatory limbo, it’s not at all clear when the U.S. government will give the all-clear for the GPT-5.6 series to go into wide release.