RAG calculator

RAG Cost Calculator

Estimate the one-time and recurring cost of a retrieval-augmented generation system. Model document ingestion, embeddings, vector storage, retrieval context, and LLM output.

Estimate RAG monthly cost

Start with a corpus scenario, replace document and query assumptions, then share or export the cost plan.

  1. 1Pick corpusChoose the closest document scenario.
  2. 2Set retrievalEnter documents, chunks, queries, and tokens.
  3. 3Check monthly costReview embedding, generation, and vector cost.
  4. 4Share or exportCopy the link or download the CSV.
Use a manual estimate for now because vector database pricing differs by provider.
Estimated monthly total$0.00
One-time embedding$0.00
Cost per query$0.00
Monthly LLM input$0.00
Monthly LLM output$0.00
Estimated chunks0

How the estimate works

RAG cost is split into indexing cost and query-time cost. The core model isembedding cost + vector database cost + LLM input cost + LLM output cost.

  • Document words are converted to estimated tokens using a default ratio of 1.33 tokens per word.
  • Chunk count increases when overlap is higher, which increases embedding and storage pressure.
  • Retrieved chunks become input context on every query, so retrieval depth drives recurring LLM cost.
Prices can change. Verify provider pricing before making purchasing or product pricing decisions.
Indexing

Estimate ingestion cost

Convert documents into chunks and embedding tokens before building the knowledge base.

Query-time

Model retrieval context

Account for retrieved chunks that become repeated input context on every answer.

Operations

Include storage and refreshes

Add vector database cost and re-indexing frequency for a more realistic RAG budget.

FAQ

What costs are included in a RAG system?

A RAG system can include one-time document embedding, recurring re-indexing, vector database storage, retrieval context tokens, and answer generation tokens.

Why does retrieval depth affect RAG cost?

Every retrieved chunk becomes input context for the language model. More chunks can improve answer coverage, but they also increase input tokens for every query.

Related calculators

Prices were last checked on 2026-07-04. Use related pages to separate indexing, token, and full product planning.