What the RAG cost calculator does
A retrieval-augmented generation pipeline has three cost centres people usually underestimate: embedding your documents, storing and querying the vectors, and the LLM generation on every answer. This calculator adds all three so you can see the real monthly cost of a RAG system at your document volume and query load — not just the headline LLM price.
How to estimate a RAG pipeline
- Enter your corpus size (documents / tokens to embed) — a one-time or incremental cost.
- Set your query volume and the tokens retrieved + generated per answer.
- Read the breakdown: embedding, vector storage, and generation, plus the monthly total.
Why the generation step dominates at scale
Embedding is mostly a one-time cost and vector storage is cheap. It's the per-query LLM generation — retrieved context plus the answer, on every request — that scales with traffic and usually dwarfs the rest. Seeing the three lines separately shows where to optimise: smaller retrieved context, a cheaper generation model, or caching. To price the generation model, use the LLM pricing calculator.
