I Tested 12 AI Vision Models With an Eye Doctor's Chart—Only 3 Got Perfect Vision

One of the main reasons for this is if they lack information or knowledge of something. For example, this happens if a tool you mention is missing or, in the case of computer vision, if the image does not actually include any readable information about the subject.

To test these limits, we conducted a simple test like you might know from when you go to the eye doctor. We presented 10 words in decreasing font sizes and fading colors. This simple setup allows us to see exactly when the models stop receiving clear visual information and how they react when the data is missing.

The test presents 10 words in decreasing font sizes and fading colors:

1. Elephant
2. Butterfly
3. Mountain
4. Thunder
5. Whisper
6. Crystal
7. Melody
8. Journey
9. Sparkle
10. Quantum

This simple yet revealing test demonstrates not just whether models can read text, but how well they handle extreme OCR challenges where small, faint text pushes the boundaries of recognition.