Embedding Model & Vector DB Cost Calculator
Enter your RAG scenario (documents, average tokens, monthly re-embedding, monthly queries) and compare every embedding model's monthly embedding cost + vector-storage cost + total in one table.
⚙️ RAG Scenario Input
Number of chunks stored in the vector DB
Average token length of one chunk
Full re-embedding runs per month (0 = one-time only)
Search requests per month
Average tokens per search query
$0.33/GB·mo
📊 Cost Comparison by Model
| Model | Dim | Context | $/1M tok | MTEB | Embed / mo | Storage | Storage / mo | Total / mo |
|---|---|---|---|---|---|---|---|---|
BGE-M3Best value BAAI · Self-hosted | 1024 | 8,192 | — | 59.5 | $0.00 | 390.63 MB | $0.1259 | $0.1259 |
text-embedding-005 Google | 768 | 2,048 | $0.006 | 62.0 | $0.3120 | 292.97 MB | $0.0944 | $0.4064 |
voyage-3.5-lite Voyage AI | 1024 | 32,000 | $0.02 | 62.0 | $1.04 | 390.63 MB | $0.1259 | $1.17 |
jina-embeddings-v3 Jina AI | 1024 | 8,192 | $0.02 | 65.0 | $1.04 | 390.63 MB | $0.1259 | $1.17 |
text-embedding-3-small OpenAI | 1536 | 8,191 | $0.02 | 62.3 | $1.04 | 585.94 MB | $0.1888 | $1.23 |
voyage-3.5 Voyage AI | 1024 | 32,000 | $0.06 | 64.0 | $3.12 | 390.63 MB | $0.1259 | $3.25 |
embed-english-v3.0 Cohere | 1024 | 512 | $0.1 | 64.5 | $5.20 | 390.63 MB | $0.1259 | $5.33 |
mistral-embed Mistral AI | 1024 | 8,000 | $0.1 | 59.0 | $5.20 | 390.63 MB | $0.1259 | $5.33 |
embed-v4.0 Cohere | 1536 | 128,000 | $0.12 | 65.0 | $6.24 | 585.94 MB | $0.1888 | $6.43 |
text-embedding-3-large OpenAI | 3072 | 8,191 | $0.13 | 64.6 | $6.76 | 1.14 GB | $0.3777 | $7.14 |
gemini-embedding-001 Google | 3072 | 2,048 | $0.15 | 68.3 | $7.80 | 1.14 GB | $0.3777 | $8.18 |
voyage-3-large Voyage AI | 1024 | 32,000 | $0.18 | 65.5 | $9.36 | 390.63 MB | $0.1259 | $9.49 |
Monthly total = monthly embedding cost (re-embed + queries) + monthly vector-storage cost. All values are reference estimates.
⚠️ Disclaimer · Data basis
All prices and MTEB scores shown are reference estimates compiled from public sources and subject to change (compiled 2026-06-20). Vector-DB cost is a simplified storage-only model and excludes read/write units, RAM, and replicas. It assumes float32 (4 bytes) and does not model int8/binary quantization savings. Confirm exact amounts in each provider's console.
Guide · FAQ · bal.pe.kr tools
- Embedding Model Comparison Guide — What embedding dimension, MTEB, context and price mean, how vector storage is calculated, and how Matryoshka reduction and vector-DB cost work.
- Frequently Asked Questions
- bal.pe.kr tools — all calculators & dev utilities
- LLM Context Window — RAG vs long-context comparison