Log response model on every Claude call to catch silent fallbacks

Pavel Espitia logs the response model in every Claude call to catch issues like silent model fallbacks and routing bugs [devto]. When using Fable 5, a prompt can silently fall back to Opus 4.8 if it trips a safeguard. Logging the response model helps identify these cases and prevent incorrect analysis or unexpected cost spikes. Espitia uses console.log("requested fable-5, served:", response.model); to log the served model [devto]. This log can help identify cases where the requested model does not match the served model, which can be due to safeguards or routing logic [devto]. Logging the response model also helps verify the effectiveness of model migrations and identify stray hardcoded model strings. By logging the response model, developers can gain insights into their LLM usage and optimize their code for better performance and cost efficiency [devto].