Every Tests OpenAI's New Astra and Sol Models on Coding, Writing and Creative Work

danshipper · x · 2026-11-18

After OpenAI released GPT-6.x Astra and GPT-6.x Sol, the Every team put both models to work on coding, writing and creative tasks, publishing a Vibe Check video and newsletter signup.

The testers include CEO Dan Shipper, senior editor Jack Cheng and head of evals Mike Taylor. The page also lists past Vibe Check headlines such as "Gemini 4 Carbon's Coding Outpaces Its Judgment" and "GPT-6 Astra Is a Big Upgrade With Some Bad Habits," hinting Astra is a clear capability jump with some behavioral quirks.

Original post →

More from Models

Models channel →