Vision model reads symbol text and answers without tools

john__allard · x · 2026-07-22

A user shows a vision-model test where the model can read a line of symbols and answer the hidden prompt without tools.

The poster says they had tried the same trick with other models a long time ago and it never worked, implying this is a notable improvement in image/text understanding rather than a generic chat reply.

Original post →

More from Models

Models channel →