Fine-tuningexplained

Fine-tuning takes a general-purpose model and trains it a little more on examples you provide, nudging it to behave a certain way — a consistent tone, a strict output format, a niche classification task it kept getting wrong. The base knowledge stays; you're specialising the behaviour.

It's easy to reach for fine-tuning too early. For giving a model access to your facts, RAG is usually the better and cheaper tool — you don't retrain anything, you just supply documents. Fine-tuning shines when the problem is how the model responds (style, structure, consistency) rather than what it knows.

The practical cost: fine-tuning needs a quality set of example inputs and outputs, and every time your requirements change you may need to redo it. For most early products, start with good prompting and RAG; consider fine-tuning once you've proven the use case and need that last mile of consistency at scale.