Fine-tuning (AI)
What is Fine-tuning (AI)?
Fine-tuning is the process of continuing to train a pre-trained language model on a smaller, curated dataset specific to a task, domain, or writing style, so the model's own weights adjust to that narrower use case instead of relying only on the general model plus a prompt. It's one of three main ways to specialize a model's behavior alongside prompting and RAG — and unlike RAG, the result is baked permanently into the model's weights rather than retrieved fresh at query time.
More detail
Fine-tuning, prompting, and RAG differ in where the specialized knowledge lives: prompting supplies context in the request itself, RAG retrieves context at query time from an external source, and fine-tuning permanently updates the model's own weights, so the effect persists across every future call without resending any source material.
Because it requires retraining, fine-tuning is a poor fit for data that changes often — pricing, inventory, policy documents — where RAG's ability to swap in updated source text at query time is faster and cheaper; fine-tuning suits teaching a model a consistent tone, output format, or specialized vocabulary that doesn't change from call to call.
Frequently asked
- What's the difference between fine-tuning and RAG?
- Fine-tuning retrains a model's own weights on a custom dataset, permanently baking in a tone, format, or specialized vocabulary; RAG instead retrieves relevant documents at the moment of a query and feeds them into the prompt, leaving the model's weights untouched. RAG is cheaper to update when source content changes often; fine-tuning suits a lasting style or skill that doesn't.
- Does fine-tuning remove the need for good prompting?
- No — a fine-tuned model still responds to whatever prompt it's given, and most production systems combine a fine-tuned or base model with careful prompting and, often, RAG for current information. Fine-tuning changes a model's default behavior; it doesn't eliminate the need to instruct it well.
See also
Build it on Orbit
Voice, messaging, email, video, and AI agents on one platform and one pay-as-you-go bill. Start free — no credit card required.