RAG supplies knowledge at query time; fine-tuning bakes behaviour into weights.
Use RAG for facts that change, are private, or need citation.
Use fine-tuning for consistent format, tone, or a specialised task.
RAG updates instantly by reindexing, while fine-tuning needs a retraining run.
The two are complementary and serious systems often use both.
The two solve different problems and are constantly confused for each other.
RAG supplies knowledge at query time by putting documents in the prompt.
Fine-tuning changes the weights so the model behaves differently.
Knowledge versus behaviour is the distinction that resolves most cases.
Use RAG when facts change, are private, or must be cited.
A policy that changes monthly belongs in an index, not in weights.
RAG updates by reindexing a document, which takes seconds.
A fine-tune updates by retraining, which takes hours and costs money.
RAG can show its sources, and a fine-tuned model cannot say where it learned something.
Use fine-tuning for consistent format, a specific tone, or a narrow repeated task.
Also use it to make a small model match a large one on that task and cut cost.
Fine-tuning is poor at teaching facts, since they are absorbed unevenly and cannot be verified.
A model that has read your documentation once is not a model that knows it.
The two combine well, with a fine-tune for behaviour and retrieval for current facts.
Try prompting first, then RAG, then fine-tuning, in that order of cost and effort.
One question on this card. Every option explains itself once you pick, right or wrong.
15 concepts in this module, then a graded test with every answer explained.