RAG Cost Calculator
Estimate the full monthly cost of a retrieval-augmented generation pipeline — indexing, retrieval, and generation.
Indexing
Query & generation
Estimated monthly cost
$74.38/month
One-time indexing
$0.30
Query embedding / mo
$0.015
Generation / mo
$74.06
Retrieved context: 1,500 tokens + 50 query tokens = 1,550 input tokens per query to GPT-5.1. Vector database hosting/storage cost is not included — add your provider's hosting fee separately.
How this is calculated
total = one-time indexing cost + (query embedding cost × queries) + LLM generation cost, where generation input tokens = retrieved chunks × average chunk size + query tokens.
Pricing is verified against provider pricing pages as of 2026-09-01. AI and cloud pricing changes frequently — confirm the current rate on the provider's own pricing page before budgeting.
Frequently asked questions
What costs make up a RAG pipeline?
Three separate costs: a one-time (or periodic) indexing cost to embed your corpus, a small per-query embedding cost for the user's question, and the LLM generation cost for producing an answer from the retrieved context — plus vector database hosting, which isn't token-based and is estimated separately.
Why does retrieved context dominate the generation cost?
Each retrieved chunk gets added to the LLM's input — retrieving 5 chunks of 300 tokens each adds 1,500 input tokens per query, on top of the user's question, often making retrieved context the largest single cost driver in a RAG system.
How can I reduce RAG costs without hurting answer quality?
Reduce chunk size or the number of chunks retrieved, use a cheaper model for generation, cache repeated system-prompt instructions, and re-index only when your corpus actually changes rather than on a fixed schedule.
Does this include vector database hosting costs?
No — vector database storage and query costs vary widely by provider (managed service vs. self-hosted) and aren't token-based, so add your vector DB's hosting fee on top of this estimate.