RAG System Cost Calculator

Estimate vector DB storage, embedding, and query costs for a Retrieval-Augmented Generation system.

Number of documents

Avg tokens per document

Chunk size (tokens)

Chunk overlap (tokens)

Embedding dimensions

Queries per day

Top-K retrieved chunks

Tab to move between fields

Total chunks

50,000

Total monthly cost ($)

$347.04

≈ 6 tanks of gas

One-time embedding cost ($)$2.56

Vector storage (GB)0.29

Vector DB monthly cost ($)$3.50

Daily query cost ($)$11.45

Monthly query cost ($)$343.54

Embedding tokens (millions)25.6

How to Use This Calculator

Enter Document Count and average Tokens per Document for your knowledge base (a page is ~500 tokens).
Set Chunk Size (512 tokens is common) and Chunk Overlap to control how documents are split for retrieval.
Enter Embedding Dimensions (1536 for OpenAI ada-002) and expected Queries per Day.
Set Top-K Retrieved Chunks to control how many context chunks are injected per LLM call.
Review Total Chunks, One-time Embedding Cost, Vector Storage (GB), Monthly Query Cost, and Total Monthly Cost to design your RAG system budget.

How the result changes with Number of documents

Number of documents: Total documents in your knowledge base corpus.
Avg tokens per document: Average token length per document. A typical page of text is ~500 tokens.
Chunk size (tokens): Size of each text chunk for embedding. Common sizes: 256, 512, 1024.
Chunk overlap (tokens): Token overlap between consecutive chunks to preserve context across boundaries.
Embedding dimensions: Vector dimensions. OpenAI ada-002: 1536, text-embedding-3-large: 3072, Cohere: 1024.
Queries per day: Expected number of RAG queries per day from your application.
Top-K retrieved chunks: Number of most-similar chunks retrieved per query for LLM context.

Total chunks: Total text chunks generated from your document corpus.
One-time embedding cost ($): Cost to embed all chunks using OpenAI ada-002 at $0.10/1M tokens.
Vector storage (GB): Estimated vector database storage in gigabytes (float32 + metadata).
Vector DB monthly cost ($): Estimated Pinecone-equivalent hosting cost (~$0.07/month per 1K vectors).
Daily query cost ($): Daily cost for query embeddings + LLM synthesis (GPT-4o pricing).
Monthly query cost ($): Projected 30-day query cost.
Total monthly cost ($): Vector DB hosting + query costs combined.
Embedding tokens (millions): Total tokens sent to the embedding API during indexing.

Identify Input Parameters
4 parameters
Number of documents = 10000, Avg tokens per document = 2000, Chunk size (tokens) = 512, Chunk overlap (tokens) = 50 = 7 input(s) provided
Calculate Total chunks
Total chunks = documentCount * chunksPerDoc
50000 = 50000
Calculate Total monthly cost
Total monthly cost = vectorDbMonthlyCost + monthlyQueryCost
347.04 = $347.04
Calculate One-time embedding cost
One-time embedding cost = (totalEmbeddingTokens * 0.10) / 1_000_000
2.56 = $2.56
Calculate Vector storage
Vector storage = totalStorageBytes / (1024 ^ 3)
0.288 = 0.288