DeepSeek V4 Lacks Vision: Model Builds Brightness Grid to Self-Correct

teortaxesTex · x · 2026-08-08

A developer tested DeepSeek-V4-flash in OpenCode to render a Lord of the Rings three.js scene. When trying to verify the output, the model realized it lacked vision capabilities.

It then demonstrated high autonomy by generating a brightness grid to check for issues and attempting to install a vision model. Upon discovering no GPU was available, the model ultimately gave up, expressing concern for the developer's laptop.

Original post →

More from Fun

Fun channel →