Every's hands-on: OpenAI's new 6.2 Astra regresses on writing and quits coding tasks too early

soumitrashukla9 · x · 2026-11-18

Dan Shipper (Every) reports that OpenAI launched 6.2 Astra and Sol 6.2, and after weeks of testing his team found the models disappointing compared to Fable 5.2:

A negative third-party review of a new flagship release, in sharp contrast to the official framing.

Related event: Every's Hands-On: Astra 6.2 Regresses in Writing, Falls Short in Coding(3 posts)→

Original post →

More from Models

Models channel →