Rumor: OpenAI's Astra uses new 'looped transformer' technique

bindureddy · x · 2026-09-02

Bindu Reddy claims that OpenAI's Astra model utilizes a new technique called a 'looped transformer.' The model loops on the same text or turn without outputting explicit chain-of-thought. This optimization is expected to make it cheaper and more performant than Fable.

Related event: OpenAI's Astra Reportedly Uses Recurrent Depth Architecture, Hiding Its Reasoning and Raising Safety Alarms(35 posts)→

Original post →

More from Models

Models channel →