Build garageTry it → read → next · ~11 min

Tutorials · Chapter D (4/4) · ~11 min

Fine-tuning lab

Try it → see it → read → next

Run a **LoRA adapter** when you need tone or format — keep **RAG** when facts change weekly.

On paths: Builder · Student

Try tools: Code & developers

Try yourself

Playground

Fine-tune vs RAG lab

LoRA for style/format; RAG for changing facts — full fine-tune is last resort.

Match brand voice on 500 support macros

Recap

What you just did

You matched tasks to LoRA vs RAG vs full fine-tune — same decision tree as the concept lesson.

Read

[LoRA](/reference/lora) starter checklist

Curate 200–500 high-quality examples
Train adapter — pin base model version
Eval vs RAG baseline on golden tasks
Serve adapter ID in prod config