Fine-Tuning vs RAG vs Prompting
Seven questions to pick fine-tuning, RAG, or prompt engineering — with honest cost ranges and anti-recommendations.
It helps you choose how to make an AI better at your specific job.
What this tool does
Prompting changes the instructions. RAG gives the model relevant information at request time. Fine-tuning changes the model’s learned behavior. This quiz helps you choose based on the problem you have.
- Choose an approach for private or frequently changing knowledge
- Decide whether a behavior problem needs examples or better instructions
- Avoid fine-tuning when a prompt or retrieval system will do the job
Start here
Sweet spot
When not to use it
Rough cost
Time to ship
Anti-recommendation
Hybrid hint
Runner-up
This is rule-based scoring, not magic. It nudges you toward the approach that matches data freshness, corpus size, volume, budget, latency, and team skills — then spells out what each path costs in time and money.
Most teams should start with prompt engineering, add RAG when facts live outside the window, and fine-tune only with evals and stable examples. The questionnaire encodes that bias on purpose.