Dictionary

Foundations

Fine-tuning

Further training a model on your own examples to change its behaviour — usually not the answer you are looking for.

Fine-tuning is often the first thing people ask about and rarely the right first step.

It teaches style and format well: a consistent tone, a specific output shape, a domain's phrasing. It teaches facts poorly, and it bakes them in at a point in time — when your data changes you must retrain.

For "the model should know about our products", RAG is almost always better: cheaper, updatable instantly, and able to cite sources.

Consider fine-tuning once retrieval and prompting are genuinely exhausted, and you have the labelled examples to do it with.

Next step

Tell us what you’re trying to build

Most engagements start with a fixed-price audit, so the first thing you buy is a decision rather than a commitment.