For me Anthropic models from 4.7 to 5 including where bad and ate tokens like crazy. Task delivery was worse than GPT 5.6 and token usage was 2-3x higher.

Looks like 5.5 is the new 4.6