Gemini 3.5 Flash aces a visual ordering puzzle with one move

iamrobotbear · x · 2026-07-29

A shared example claims Gemini 3.5 Flash is very strong at image reasoning. The screenshot shows the model solving a visual ordering puzzle: five numbered pipes need to be rearranged into ascending order from top to bottom, and the model answers that only one pipe must be moved.

The post emphasizes that this is not just object labeling or OCR. It is framed as a case where the model has to infer the structure of the scene, reason through the ordering, and produce the minimal move count.

Original post →

More from Models

Models channel →