Your question is Debugging LLM Latency. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How would you debug a high-latency issue in a deployed LLM service?
Explain the measurements, instrumentation, and controlled experiments you would use to separate client or network delay from Baseten deployment queueing, cold starts, model prefill, token generation, and response streaming. Keep the answer focused on a production debugging workflow rather than model-quality analysis.