How accurate are these estimates?

They are planning-grade approximations using published list pricing and typical sizing assumptions (4 bytes per float dimension plus index overhead). Real bills vary with replicas, pod sizes, free tiers, and negotiated rates, so treat the numbers as a relative comparison rather than an invoice.

How is storage size calculated?

Raw vector size is vectors × dimensions × 4 bytes (float32), and the tool adds roughly a 30% overhead for the index structure and metadata. That total drives the storage component of each provider's cost.

Why is pgvector so much cheaper?

pgvector runs inside Postgres, so its cost is just the server you already pay for — the estimate uses a flat managed-Postgres instance fee. The trade-off is that you operate it yourself: scaling, backups, and index tuning are on you, unlike the fully managed services.

Does it call any pricing API?

No. The pricing models are built in and evaluated locally in your browser. Update the numbers as providers change their list prices.

Vector Database Cost Comparison

Vector database cost comparison

Choosing where to store your embeddings is partly a feature decision and partly a bill. This tool estimates the monthly cost of the same workload across four popular options — Pinecone, Weaviate Cloud, Qdrant Cloud, and self-hosted pgvector — from three inputs: how many vectors you store, their dimensionality, and how many queries you run per month.

How it works

First the tool computes raw storage as vectors × dimensions × 4 bytes for float32 embeddings, then adds about 30% for index and metadata overhead. Each provider applies its own model on top of that storage figure: a managed serverless service charges per gigabyte stored plus a per-query rate; pod-based pricing maps the storage tier to an instance fee; and pgvector is modelled as a flat managed Postgres instance whose cost barely moves with scale until you outgrow the box. The side-by-side table then ranks the monthly totals so you can see where the break-even points fall as your data and traffic grow.

Tips and notes

Dimensions dominate storage. Halving from 1536 to 768 dimensions nearly halves storage cost — quantization or smaller embedding models pay off fast at scale.
Query volume matters for serverless. Per-query pricing rewards spiky, low-traffic apps and punishes high-QPS workloads, where a pod or self-hosted setup wins.
pgvector is cheapest until it isn’t. It is the budget option for millions of vectors, but past tens of millions the index maintenance and recall trade-offs push teams toward purpose-built engines.
Estimates are local and editable. Refresh the inputs whenever provider list prices change.