RAG Cost Calculator

One-off indexing cost and monthly running cost of a RAG pipeline

FreeNo credit cardResults in 2 minutesUsed by 500+ companies

1. Your documents

2. Your queries

Result

One-off: index your documents
$0.77
Per month
$55
Per query
$0.0055
Index size
0.143 GB
  • 20,000 chunks of 7,152 bytes each.
  • One-off: embedding $0.20 + Pinecone writes $0.57 (143,040 write units).
  • Monthly Pinecone: $50 (usage $0.09: storage $0.05 + reads $0.04, billed at the $50 minimum when lower).
  • Monthly model: answers $4.50 (3,000 input tokens per query) + query embeddings $0.01.

The formulas

Chunks          = document tokens / tokens per chunk (rounded up)
Record size     = 8-byte ID + metadata bytes + dimensions x 4 bytes
Index size (GB) = chunks x record size / 1,000,000,000
One-off         = document tokens / 1M x embedding price + write units / 1M x $4   (1 write unit per KB)
Pinecone/month  = max($50, index GB x $0.33 + queries x max(0.25, index GB) read units / 1M x $16)
Answers/month   = queries x ((top-k x chunk tokens + prompt tokens) x input price + output tokens x output price) / 1M
Per month       = Pinecone + answers + query embeddings
What this number does not include. Re-indexing when documents change, reranking, hosting the app that runs the pipeline, Voyage's free first 200 million tokens, prompt caching, the build itself, and Pinecone prices at the top of its published ranges (this uses the low end). Token counts are your assumptions, so measure a few real queries.

How this works

A RAG system splits your documents into chunks, turns each chunk into a vector with an embedding model, stores the vectors, and at query time finds the closest chunks and sends them to a language model with the question. The calculator prices each of those steps with published rates.

Vector storage uses Pinecone's own formulas. A record is an 8-byte ID plus its metadata plus 4 bytes per vector dimension. Writes use 1 write unit per KB. A query uses 1 read unit per GB of index, with a minimum of 0.25. Pinecone Standard lists storage at $0.33 per GB a month, write units at $4 to $4.50 per million and read units at $16 to $18 per million, with a $50 monthly minimum; the calculator uses the low end. Starter is free up to 2 GB, 2 million write units and 1 million read units a month.

Worked example with the defaults: 10,000,000 / 500 = 20,000 chunks of 8 + 1,000 + 1,536 x 4 = 7,152 bytes, so a 0.143 GB index. One-off: 10 x $0.02 = $0.20 to embed plus 143,040 write units x $4 per million = $0.57. Monthly: Pinecone usage is $0.05 storage + 2,500 read units ($0.04), so the $50 minimum applies. Answers on Claude Haiku 5.5 are 10,000 x (3,000 x $0.10 + 300 x $0.50) / 1,000,000 = $4.50, and query embeddings $0.01. Total $54.51 a month.

Sources (read November 11, 2026)

Need a RAG pipeline that answers accurately from your own documents? See RAG pipeline development.

Frequently Asked Questions

Related Tools

AI Chatbot Cost Calculator

Monthly cost of a support chatbot against per-resolution pricing.

Try it free

AI Agent Cost Calculator

Monthly and per-run cost of an AI agent.

Try it free

AI Agent Development Cost Estimator

Price a build from your hour estimates.

Try it free

Want expert help implementing this?

Our automation specialists can help you achieve these results and more. Book a free consultation to discuss your specific needs. Or see the exact service this tool maps to: RAG Pipeline Development.

Book a Free Consultation