Analysis suggests OpenAI is testing an unreleased internal model

AccBalanced · x · 2026-08-19

Analysis indicates that OpenAI has been testing an internal model since at least early May that has never been publicly released.

In contrast, Astra's designation as "Critical" for cybersecurity was announced less than two weeks ago, implying its testing occurred around July-August. This discrepancy suggests these are two different models. The analysis cites Sam Altman's statement that they "still expect to ship great new models soon," implying Astra is a further-out release while new models are imminent.

Related event: OpenAI's Next-Gen Model 'Astra' Rumored Imminent with AGI-Level Claims(9 posts)→

Original post →

More from Companies & People

Companies & People channel →