Your question is Centralized vs Task-Specific LLMs. 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).
You are deciding how to power several NLP features in one product, including classification, extraction, and question answering. The team can either deploy one large centralized language model or maintain multiple smaller models tuned for each task.
Discuss the tradeoffs between deploying a large centralized LLM versus multiple smaller, task-specific models.