Fine-tuning
Fine-tuning is training an existing AI model further on your own examples, so it adapts its behaviour to a particular task, style or domain. It changes the model itself, unlike retrieval, which only changes what the model is given to read.
What it means
Fine-tuning helps when a model needs to behave differently, for example following a specific format or a house style, across many requests. It's less useful for giving a model new facts, which retrieval usually does better and more cheaply.
It also needs enough good examples, and a way to evaluate whether the tuned model is actually better, which many teams discover only after they've paid for the training.
How we use it
Illustrative example In the illustrative clinic brief on our Cut It page, a custom-trained model is the feature we'd never build: an off-the-shelf model does the job, and a custom one costs more, needs data the clinic doesn't have, and changes nothing a patient notices.
Related terms
Published 28 September 2026. All terms