The deltas between the others are mostly not significant either. They're all about equally good. There's no categorical difference between GPT-4 and Claude 3.5.