Fine-tuning further trains a base model on your own labeled examples to specialize its behavior, tone, or output format. It complements - rather than replaces - prompting and evaluation: you still need datasets and evaluators to prove a fine-tuned model is better and to catch regressions.
Why it matters
Fine-tuning can make a smaller, cheaper model match a larger one on a narrow task - a direct way to cut cost and latency once you know the task well.
How it works
You train a base model further on your own labeled examples. Because it depends on data quality, fine-tuning is inseparable from datasets and evaluation: you need a golden set to prove the tuned model is actually better and to catch regressions.