All tools
AI toolkit
RAG sizing & cost
Plan a retrieval system end to end. Inputs are document count, average length, chunk size and overlap, and embedding model. Outputs are chunks, vector DB size at float32, one-time embedding cost, and a monthly cost estimate across three popular hosts.
Corpus
Chunking
Effective stride: 448 tokens per chunk → 3 chunks per document.
Embedding model
Sizing
- Total chunks
- 15,000
- Total tokens to embed
- 7.68M
- Vector DB size (float32)
- 91.6 MB
- One-time embedding cost
- $0.1536
Estimated monthly storage cost
- Pinecone Serverless
- $0.0295 /mo
- pgvector (self-hosted)
- $6.00 /mo
- Chroma (self-hosted)
- $6.00 /mo
Pinecone uses ~$0.33 / GB-month storage; reads/writes are billed separately and not included here. Self-hosted estimates assume a small VPS as a floor.