Choose Between LLM API and Fine-Tuning

Scenario

You are building an NLP feature for a product that needs to answer or classify user text with consistent quality. Your team is deciding between calling a foundational LLM API and hosting a smaller open-source model that you fine-tune on your own data. The choice affects accuracy, latency, cost, privacy, and how much control you have over model behavior.

Question

Explain the trade-offs between using a foundational LLM API versus hosting a smaller, fine-tuned open-source model.