About · Data Sources · Methodology
Last updated 2026-06-20 · This page documents the tool's trust (E-E-A-T) information.
What this tool is
The Embedding Model & Vector DB Cost Calculator compares major text embedding models — OpenAI, Voyage AI, Cohere, Jina, Google, BGE and Mistral — by dimension, maximum context, per-1M-tokens price and MTEB score in one table. Enter a RAG scenario (document count, average tokens, monthly re-embedding and monthly queries) and it instantly estimates each model's monthly embedding cost, vector storage cost and total. No signup is required and all calculations run entirely in your browser.
We built it because embedding cost can differ by more than 20× per token and dimension drives storage cost, yet few interactive tools combine both into a single monthly total. You can also simulate how much Matryoshka dimension reduction saves on storage.
Data sources
All model specs, prices and MTEB scores are compiled from the public sources below. Data was compiled on 2026-06-20.
- OpenAI — Embeddings & pricing
- Voyage AI — Pricing
- Cohere — Pricing
- Jina AI — Embeddings
- Google Vertex AI — Generative AI pricing
- BAAI BGE-M3 (Hugging Face)
- MTEB Leaderboard
- Pinecone — Pricing
- Qdrant — Pricing
- pgvector
Methodology
- Embedding cost = (total tokens ÷ 1,000,000) × price per 1M. Total tokens = documents × average tokens.
- Monthly cost = one-time index cost × re-embeds per month + monthly query-embedding cost.
- Storage size = dimension × documents × 4 bytes (float32). GB conversion uses 1 GB = 1024³ bytes.
- Storage cost = storage (GB) × storage rate ($/GB·month).
Limitations
- All prices and MTEB scores are reference estimates and change frequently with provider policy.
- Vector-DB cost is a storage-only model; it excludes read/write units, RAM, replicas and minimums.
- It assumes float32 (4 bytes) and does not model int8/binary quantization (4×–32× savings).
- MTEB is an English-centric average and does not guarantee performance for a specific language or domain.
Privacy
Your inputs are never sent to a server; they are stored only in your browser (localStorage) and the share-link URL token (?s=). No account or tracking cookie is used — only ads and access analytics needed to run the site are collected.
Operator
This tool is built and operated by bal.pe.kr, a Korean solo micro-SaaS portfolio. The goal is to help developers and RAG engineers quickly compare embedding models from a cost perspective.
- Operator: bal.pe.kr (solo-run micro-SaaS)
- Contact: comsamo84@gmail.com
- More tools: bal.pe.kr tools
Disclaimer
All prices, scores and costs provided are reference estimates; actual billing and performance vary by provider policy, model version and usage. Always verify against each provider's official pricing and your own evaluation set before committing.