Tutorials · Chapter C (3/4) · ~9 min

Fine-tuning vs RAG

Try it → see it → read → next

Change the model’s habits (fine-tune) or hand it open-book notes (RAG).

Try yourself

Playground

Pick the lever

Prompt, RAG, or fine-tune? One scene at a time.

Answer from our company wiki that changes weekly

Which tool fits this job?

Recap

What you just did

You walked a RAG pipeline. Fine-tuning is a different lever.

Teach

How it works

See it

Change the model vs give it notes

Fine-tune

Teach voice / format into weights

RAG

Fetch fresh docs at ask time

Fine-tune = bake in style · RAG = look things up when answering

| Approach | What changes | Best for | | --- | --- | --- | | Prompting | Instructions only | Quick style/tasks | | RAG | Retrieved docs at ask-time | Fresh or private facts | | Fine-tuning | Model weights on your examples | Style, format, domain language |

Fine-tuning won’t magically stay up to date with next week’s inventory. RAG can pull today’s doc. Fine-tuning can make the model sound like your brand or follow a schema more reliably.

Practical order: prompts → RAG → fine-tune if still stuck.

Use it

When you'd use this

Company wiki Q&A → RAG first.
“Always reply as JSON with these fields” at scale → fine-tune or strong prompting + validation.

Watch out

Fine-tunes cost money and can forget old skills (drift). RAG fails if retrieval misses the right chunk.

Try next

Try this next

For a school handbook bot, pick RAG or fine-tune and say why in one sentence.