Bedrock Knowledge Base Cost Calculator
Compare S3 Vectors vs OpenSearch Serverless retrieval costs for your corpus — see the minimum OCU floor before you commit.
- No signup
- Instant results
- ~3 min
Rates reviewed
Read the full breakdown — Companion pricing guide covers billing dimensions, traps, and a 30-day cleanup plan.
Your document corpus
us-east-1list rates, pricing as of 2026-07-06. Knowledge Bases have no separate fee — you pay embeddings, vector store, and model inference. Aurora pgvector not modeled. Source: AWS public pricing.
PDFs and wikis average 8–32 KB of extractable text.
Who This Tool Is For
CTOs, GenAI platform teams, and FinOps leads sizing retrieval-augmented generation before production commit.
Why We Built This Tool
OpenSearch Serverless carries a ~$350/month OCU minimum even when idle — teams underestimate retrieval-layer cost before the first Knowledge Base ships.
What Problem It Solves
- OpenSearch minimum floor. Production collections require 2 OCUs at $0.24/OCU-hour — roughly $350/month with zero queries.
- Embedding re-ingestion. Daily sync on a large corpus compounds Titan embedding charges every month.
- Inference dominance. At scale, model tokens often exceed vector storage — but the floor still matters for pilots.
See our RAG on Bedrock pattern or the Bedrock pricing calculator.
Frequently Asked Questions
Is there a separate Knowledge Bases fee?
No. You pay for embeddings, the vector store backend, optional parsing/reranking, and foundation model inference. There is no per-KB control-plane charge.
When should I use S3 Vectors vs OpenSearch Serverless?
S3 Vectors is the 2026 default for cost-sensitive RAG where 50–200ms retrieval latency is acceptable. OpenSearch Serverless fits hybrid BM25+vector search and sub-50ms SLAs — but carries a ~$350/month minimum.
What embedding model does this assume?
Amazon Titan Text Embeddings v2 at $0.02 per 1M tokens. Custom embedding models bill at their respective Bedrock rates.
Does this include source document S3 storage?
No. Source documents in S3 bill at standard S3 rates separately. This calculator covers KB ingestion embeddings and vector index storage.
How accurate is the OpenSearch estimate?
We model the production 2-OCU minimum floor plus managed storage. Actual OCU scaling depends on indexing and query load — use the OpenSearch calculator for detailed OCU sizing.
Where do these prices come from?
AWS published us-east-1 list prices (Bedrock pricing), pricing as of 2026-07-06. S3 Vectors rates follow AWS blog workload math at corpus scale.