Does http://B.AI secretly switch your requests to other models when popular models are overloaded? For example, you select Claude Opus 5.5 in Claude Code via the http://B.AI API, but notice that some of your usage records show Sonnet 5. What’s going on? Seeing different models in your usage records doesn’t necessarily mean http://B.AI secretly switched your model. Claude Code’s subagent tasks, model configurations, and fallback mechanisms can all lead to different models being used within the same workflow. Here’s a breakdown of the possible causes and how to investigate them 🧵👇