How it works
You prepare pairs of example inputs and ideal outputs, from dozens to thousands of them, and the provider (or your own GPUs) trains the model on them for a few passes, called epochs. Full fine-tuning updates all the weights; parameter-efficient methods such as LoRA train a small add-on instead, which is far cheaper and is how most open models are tuned. Preference and reinforcement tuning go further by rewarding better answers over worse ones.
Fine-tuning is good at behaviour: a consistent tone, a strict output format, sorting tickets into your categories, or getting a small, cheap model to match a bigger one on one task. It is a poor way to teach facts that change, which is RAG's job. A tuned model is tied to its base model and must be retrained when that model is retired. Availability shifts too: OpenAI closed its self-serve fine-tuning to new customers in 2026, while Google Cloud still offers it and open models can be tuned on your own GPUs.
Fine-tuning pros and cons
Pros
- Locks in a tone, format or task more reliably than prompting
- Shorter prompts, since the instructions live in the model
- Lets a small, cheap model handle one narrow job well
Cons
- Needs a good set of example data, which takes effort to build
- Training costs money, and tuned models may cost more to run
- Must be redone when the base model is retired
- Does not reliably teach new facts
When to use Fine-tuning
Pick it when
- A prompt with examples still gives inconsistent format or tone
- High-volume, narrow tasks where a small tuned model saves money
- You have hundreds of good examples of the output you want
Skip it when
- The goal is answering from documents (use RAG)
- Better prompts or structured output have not been tried yet
Fine-tuning pricing
Pay as you go
Charged per training token (data size times passes) plus use. On Google Cloud about $1.50 to $25 per million training tokens; tuned newer Gemini models cost 1.5 times the base rate to use.
Fine-tuning pricing page (opens in a new tab)Approximate, checked September 2026.What the other tools cost
Fine-tuning vs the alternatives
Related terms
More in AI and LLMs
Building with models