Simple letter-counting test exposes huge gap: GPT-6-Astra hits 93%, Fable 5.1 flounders

scaling01 · x · 2026-09-18

Original post →

More from Models

Models channel →