New Vision in ChatGPT is worse

Share examples here of the new GPT-4-Vision model available in ChatGPT to Plus subscribers.

From my testing, it seems worse at understanding concepts and recognizing fictional characters.

I have some other examples that don’t involve copyrighted characters, but this is one of the more clear comparisons I have. Of course the Old model wasn’t perfect, but it did try.

Oddly enough, If you are the one to identify it, it has no problem recognizing the characters, and seems to be actually better at it than the old model. So perhaps it’s some arbitrary restriction?