← AI glossary

Definition

Fine-Tuning

Fine-tuning adapts an existing AI model on task-specific examples so it behaves better for a defined use case.

Also known as: model fine-tuning, finetuning

Short definition

Fine-tuning is the process of taking a pretrained model and training it further on a smaller, more specific dataset. The goal is to improve behavior for a domain, format, tone or task that generic prompting does not handle well enough.

How it works

Teams prepare high-quality examples, train the model for additional steps and evaluate whether the adapted model improves over the base model. Fine-tuning changes model behavior more deeply than a prompt, so mistakes in the training data can become persistent.

Example

A company might fine-tune a model to classify support tickets using its own internal categories and writing conventions. This can make outputs more consistent than asking a general model through prompts alone.

Why it matters

Fine-tuning can improve consistency, reduce prompt length and specialize a model. It is not always necessary: retrieval, better prompts or structured rules are often simpler. The decision should be based on measured performance, not hype.