모션 디자인 벤치마크

게시물 내용

Kimi K3.1 on the usual "motion design" benchmark. Taste is better than deepseek and gpt 6 sol I'd say, lower than Sonnet and Opus. And honestly everyday that goes by "taste" ie "can it figure out what I want done" is becoming more important than "can it do the thing I specified"

— @mesmerlord