Home › Solutions › RAG Chatbot: Cost per Query
Use caseRAG Chatbot: Cost per Query
In RAG, retrieved context is usually most of the input. How many chunks you retrieve, and how big they are, drives the bill.
Cost per query–
Monthly cost–
Yearly cost–
Saved at 20% (our guaranteed minimum, after fee)–
Defaults use GPT-6.1 Sol's list prices ($2 input, $0.10 cached, $10 output per 1M tokens, checked October 30, 2026). Enter your own model's prices.
Where the cost goes
- Retrieved chunks, often more than the answer needs.
- The system prompt and instructions.
- The answer itself.
- Embedding the query (small) and re-embedding documents (occasional).
How to cut it
- Lower top-k and chunk size until quality drops, then step back.
- Cache the system prompt.
- Rerank to send fewer, better chunks.
- Semantic caching for repeated questions.
Frequently asked questions
How much does RAG cost per query?
Retrieved and prompt tokens times the input price, plus output tokens times the output price, plus a small embedding cost.
How do I reduce RAG costs?
Retrieve fewer, better chunks, cache the system prompt, and cache answers to repeated questions where that's safe.
Want us to find your 20%?
Book a free 15-minute API cost teardown. Bring last month's provider usage and we'll show you your biggest waste and the first fix we'd ship.
Get a Free 15-Minute API Cost Teardown