Sonnet 5 was supposed to be cheaper. It cost me more than Fable 5

Tested Claude Sonnet 5 and Fable 5 on two coding tasks. One was a RAG Debugger added inside the 400K line Open WebUI repo. Other was a full browser based Clash Royale style game. First, I asked both models to add a RAG Debugger to Open WebUI, which is a codebase with around 400K lines. The debugger had to show: document processing status extracted text generated chunks chunk metadata retrieval results source details retrieval scores Both models did a good job. Fable built a debugger that worked across the f