Back to Learn
Production
Advanced
9 min read

Fine-Tuning vs Prompting vs RAG

When to reach for which tool.

Fine-Tuning vs Prompting vs RAG

A common (and expensive) mistake: jumping to fine-tuning before exhausting cheaper options. Decision tree:

Does the model already know the answer?
├── Yes → Better prompting
└── No  → Does it need facts that change?
         ├── Yes → RAG
         └── No  → Does it need a new skill or style?
                  ├── Yes → Fine-tuning (LoRA first, full only if needed)
                  └── No  → Reconsider — maybe a different base model

Rough costs

ApproachSetupPer-requestMaintenance
Prompting$0Inference onlyNone
RAG$50-$500Inference + retrievalRe-index when data changes
Fine-tuning (LoRA)$20-$2,000Inference (slightly higher)Re-tune when base model deprecates

A pragmatic order

  1. Try harder on the prompt. A great prompt closes 70% of the gap.
  2. Add RAG. If the gap is factual, retrieval almost always wins.
  3. Switch base models. Sometimes a bigger or different model already does what you want.
  4. Then consider fine-tuning. And start with LoRA — full fine-tuning is rarely worth it.
LLMAtlas — The Open Ecosystem Workspace for LLMs