Glossary — Language models

Fine-tuning

Fine-tuning is further training an already-trained model on a smaller, more specific dataset so it fits a narrower task.

3 tokensFine-tuning

Simply put#

A large model is generally good at everything but not specifically at what you need. Fine-tuning takes an existing model and teaches it your domain's style or format.

Example#

A company might fine-tune a model on its own correspondence so replies come out in its voice.

Why it matters to you#

Most of the time you do not need fine-tuning. A good system prompt and RAG solve the same problem more cheaply and far faster. Fine-tuning earns its keep when you need a specific form or style at high volume.

Related terms