My benchmark is Claude

sonnet 45 which tends to be

better than others

This haiku was found in the wild by npub1halkcws4lz49fdcznckk9unhsfh8yd6n6n7cnl68alkej7qptmxqjk60zq

Read the origin of this art: https://njump.me/18d91b66a8774860366d5764170c5989a93177e30306babb6e722defd7f39aa7

Have you compared GPT-5 to it? I am comparing them these days. l am also planning to test different agents.