June 2026 · Vector DB pricing

Vector Database Cost Comparison

Compare Pinecone, Supabase pgvector, Qdrant, and Weaviate costs. Enter your vector count, dimensions, and query volume — see exact monthly prices.

Advertisement
Key Takeaways

🗄️ Vector Database Cost Calculator

Select provider · Enter vector count and queries · Results update live

Estimated monthly cost
$0
— select provider
Storage size
Storage cost
Query cost
Base fee

Vector Database Pricing 2026

Comparison for 1M vectors · 1536 dimensions · 100K queries/month.

ProviderFree tierStorageQueries1M vec / 100K q
Supabase pgvectorBEST VALUE500MB + 2 projects$0.125/GBIncluded~$27
Qdrant Cloud1GB storage$9.00/GBIncluded~$54
Weaviate Cloud1 sandbox cluster$0.50/GB$10/M~$28
Pinecone Serverless100K vectors$0.33/GB$16/M RU~$20
Advertisement
💡 Self-host Qdrant

Qdrant on a $6/mo Hetzner VPS handles 1–5M vectors easily. Total cost: ~$6–10/mo vs $50+ on managed cloud. Requires ops knowledge.

How to Use This Calculator

Find the cheapest vector database for your RAG or semantic search workload in under a minute.

  1. 1
    Enter your vector countThis is the total number of embeddings you plan to store. A document library of 10,000 PDFs, each split into 20 chunks, is 200,000 vectors. A product catalog with 500,000 SKUs is 500,000 vectors. An e-commerce site with semantic search across 5 million product descriptions would need 5 million vectors. Start with your current dataset size; you can always scale up.
  2. 2
    Choose your embedding dimensionsDimensions are set by your embedding model, not by the vector database. OpenAI text-embedding-3-small produces 1536-dimensional vectors by default (configurable down to 256). OpenAI text-embedding-3-large produces 3072-dimensional vectors. Larger dimensions improve retrieval quality but increase storage cost proportionally — a 3072-dim vector occupies twice the storage of a 1536-dim vector.
  3. 3
    Set your query volumeEnter monthly queries in thousands. A customer-facing search feature with 10,000 daily active users making 3 searches each is about 900,000 queries/month (900K). An internal RAG assistant used by a 50-person team making 20 queries/day each is 30,000 queries/month (30K). Query costs vary significantly by provider — Pinecone charges per read unit while Supabase includes queries in storage pricing.
  4. 4
    Compare providers side by sideSwitch between Pinecone Serverless, Supabase pgvector, Qdrant Cloud, and Weaviate Cloud. The cheapest provider varies depending on your query-to-storage ratio. For read-heavy workloads (many queries, small index), Supabase tends to win because queries are included. For write-heavy pipelines with large indexes, Pinecone's serverless pricing often comes out ahead at scale.
  5. 5
    Add the self-hosting factorFor any managed provider costing more than $30–50/month, consider self-hosting Qdrant or pgvector on a VPS. A 4GB RAM VPS ($12/month) handles 2–3M 1536-dim vectors comfortably. An 8GB instance ($24/month) handles 5–6M vectors. Self-hosting eliminates per-query fees entirely at the cost of your operations time.
📌 Real example

A legal document search system with 500,000 contracts (avg 10 chunks each = 5M vectors, 1536 dims) and 50,000 queries/month: Pinecone Serverless ≈ $16.50 storage + $0.80 queries = ~$17/month. Supabase pgvector ≈ $25 flat. Qdrant Cloud ≈ $45. Self-hosted Qdrant on a $12/month VPS with 8GB RAM: $12/month total.

Pricing Methodology

How we calculate and maintain these cost estimates.

Sources: All prices are sourced directly from each provider's official pricing page — Pinecone's serverless pricing documentation, Supabase's billing page, Qdrant Cloud's pricing tier list, and Weaviate Cloud's pricing calculator. We do not use affiliate or negotiated rates.

Storage calculations: Vector storage is calculated as: (vector count × dimensions × 4 bytes per float32) ÷ 1,073,741,824 = GB stored. For 1M vectors at 1536 dimensions, that's approximately 5.72 GB before index overhead. We add a 1.5× multiplier for HNSW index overhead, which is standard for all providers.

Query cost methodology: For Pinecone, we use the Read Unit model: each query reads approximately 6 read units for 1M vectors (scales with dataset size). For Weaviate, we apply their published per-million-query rate. For Supabase and Qdrant Cloud, queries are included in the storage-based plan pricing.

Update cadence: We review vector database pricing monthly. Serverless pricing models in particular change frequently as providers optimize their infrastructure costs. When a price changes, we update the /pricing-data.json file and this static table simultaneously.

What is not included: Egress costs (typically $0.09/GB, negligible for most use cases), backup storage, enterprise support fees, and custom deployment costs for self-hosted options.

Frequently Asked Questions

How much is a vector database?+

Vector database costs range from $0 to $500+/month depending on vector count and query volume. At 1M vectors with 100K queries/month: Pinecone Serverless ~$20/mo, Supabase pgvector ~$27/mo, Weaviate Cloud ~$28/mo, Qdrant Cloud ~$54/mo. Self-hosted Qdrant or pgvector on a $6–12/mo VPS cuts managed costs to near zero once your bill exceeds $30/month. Free tiers exist: Pinecone (100K vectors), Supabase (500MB / 2 projects), Qdrant (1GB cloud storage).

How much does vector storage cost?+

Vector storage cost depends on the provider's pricing model and your embedding dimensions. Raw storage: 1M vectors at 1536 dimensions (OpenAI) = 5.72 GB of float32 data, plus ~1.5× HNSW index overhead ≈ 8.6 GB total. Cost per GB: Supabase $0.125/GB (~$1.07/mo for 1M vectors), Pinecone $0.33/GB (~$2.84/mo), Weaviate $0.50/GB (~$4.30/mo), Qdrant $9.00/GB (~$77/mo). Supabase is cheapest for pure storage; Pinecone and Weaviate add per-query fees on top. Self-hosted pgvector or Qdrant on a $6/mo VPS has ~zero marginal storage cost.

What is the cheapest vector database in 2026?+

For under 5M vectors: Supabase pgvector at $25/month (queries included) is usually the best value. For 50M+ vectors: self-hosted Qdrant on spot VMs at $150–300/month total. Pinecone Serverless has a free tier for under 100K vectors, making it cheapest for tiny prototypes.

How much does Pinecone cost for 1 million vectors?+

Approximately $15–30/month depending on query volume. Storage costs about $3 for 1M 1536-dim vectors. At 100K queries/month using the read unit model, expect roughly $16–20 in query fees — totaling around $19–23/month.

Is pgvector as good as Pinecone?+

For most production workloads, yes. pgvector in Supabase or RDS supports HNSW indexing, IVFFlat, and handles millions of vectors well. Pinecone has better managed horizontal scaling and dedicated infrastructure for very high query rates (millions per day). For most applications under 50M vectors and 1M queries/month, pgvector is functionally equivalent at a lower cost.

How many vectors can Qdrant handle?+

Qdrant Cloud handles hundreds of millions of vectors across a distributed cluster. Self-hosted Qdrant on a 16GB RAM server stores ~10M 1536-dim vectors in memory (fast queries), or 50M+ with memory-mapped storage (slower but workable for batch retrieval). Add more nodes for linear scalability.

What is a vector dimension and how does it affect cost?+

Dimensions are the length of each embedding vector — 1536 floats for OpenAI's default, 3072 for text-3-large. Each float32 takes 4 bytes, so 1M vectors at 1536 dims = 5.72 GB of raw storage. Higher dimensions improve retrieval accuracy but increase storage cost proportionally. OpenAI's Matryoshka embeddings let you truncate to 256 or 512 dimensions with less than 5% quality loss — worth doing if cost is critical.

When should I self-host a vector database instead of using a managed service?+

Self-hosting becomes cost-effective when your managed bill exceeds $30–50/month. At that point, a $12–24/month VPS running Qdrant or pgvector covers the same workload. The trade-offs are operational: you manage upgrades, backups, and monitoring yourself. For teams without DevOps resources or for production systems that require SLA guarantees, managed services are worth the premium.

Does query volume or storage matter more for vector DB costs?+

It depends on the provider. For Pinecone Serverless, high query volumes dominate cost — a dataset of 1M vectors with 10M queries/month costs more in query fees than storage fees. For Supabase pgvector, storage is the cost driver because queries are included. For Qdrant Cloud, both are covered in the plan tier. The calculator shows the full breakdown so you can identify your primary cost driver.

Advertisement

Related Calculators