Skip to content
Module
AI

What is fine-tuning? LLM fine-tuning for PMs

Fine-tuning means continuing the training of an already trained model on a set of examples specific to one task: question and answer pairs, texts in the desired style, corrected labels. The model's parameters change, which alters its behavior durably. The result is a specialized version of the base model that then has to be hosted, evaluated and maintained.

Definition written by Alex Dionisio, Senior Product Manager. Updated November 13, 2026.

Why it matters for a PM

Fine-tuning often comes up as the way to "teach the model our data." In practice it mostly locks in a format, a tone or a repetitive task, and it handles changing knowledge poorly. A PM who knows this avoids a long and expensive project in cases where a better prompt, well chosen examples or RAG would have done the job.

Example

A hiring platform sorts thousands of job ads by occupation. With a detailed prompt, a large model gets the labels right, but every call is expensive. The team labels a few thousand ads and fine-tunes a smaller model on them. On the test set, the tuned model comes close in quality at a much lower cost per call.

Key points

  • Fine-tuning shapes behavior and style; for current facts, RAG is usually the better fit.
  • It needs a clean, representative set of examples, which is often the heaviest part of the work.
  • A tuned model has to be evaluated again whenever the base model changes.
  • It can sometimes replace a large model with a smaller one on a narrow task.

Common mistakes

  • Jumping to fine-tuning before getting the most out of the prompt and in-prompt examples.
  • Training on personal data without a legal basis or the vendor's terms allowing it.
  • Ignoring upkeep: refreshing data, retraining and ongoing evaluation.

Go further with Module

The courses and lessons that cover this concept:

Guide on this topic