Simple pricing
Priced per million vectors stored. No per-seat charge, no charge for queries.
For prototypes and side projects
- 1M vectors, 1 index
- Shared region, 50 queries/sec
- Hybrid search and filters
- Community support
For teams with production traffic
- 100M vectors included, then $0.08 per million
- Dedicated region, 5,000 queries/sec
- 24 ms p99 and 99% recall@10 targets
- Weekly recall report
- 99.9% uptime SLA, shared Slack channel
For regulated and very large indexes
- Everything in Scale
- Billions of vectors, no ceiling
- VPC peering or run-in-your-account
- 99.99% uptime SLA
- Named engineer and data residency
- Audit logs and SSO
Compare plans
Frequently Asked Questions
How is a vector counted?
By what is stored, measured hourly and averaged over the month. Dimensions do not change the price; a 384-dimension vector and a 3,072-dimension vector cost the same. Queries, filters and rebuilds are free.
What happens if you miss the latency target?
The weekly report shows the measured p99 on your own traffic. Two consecutive weeks above the target credits that month back. We would rather refund than argue about a benchmark.
Can we bring our own embedding model?
Yes. Send vectors you have already computed, or point us at a model endpoint and we will embed on write. Nodeform stores and searches; it does not lock you to a model.
Can we run it in our own account?
On Enterprise, yes. The data plane runs inside your VPC on AWS, GCP or Azure and we operate it from the outside. Same API, your network, your keys.
Ready to stop running
your own vector index?
Read the quickstart, or talk to the engineer who will run your index.
