May 14, 2026 · Qdrant
Qdrant: Open-Source Vector Search Engine
Qdrant is an open-source vector search engine written in Rust. It runs anywhere: your laptop, a single cloud VM, a Kubernetes cluster, or Qdrant Cloud (managed). The MIT license means no per-seat fees
Overview
Qdrant is an open-source vector search engine written in Rust. It runs anywhere: your laptop, a single cloud VM, a Kubernetes cluster, or Qdrant Cloud (managed). The MIT license means no per-seat fees and no restrictions on self-hosted production use.
Pricing
Self-hosted: Free. Run on any infrastructure.
Qdrant Cloud:
| Tier | Specs | Cost |
|---|---|---|
| Free | 1GB RAM, 4GB disk | $0 |
| Standard | Dedicated cluster | $0.078/GB-hour (~$57/month per GB RAM) |
| Hybrid Cloud | Your infra, Qdrant management | Contact sales |
| Private Cloud | Full on-premise | Contact sales |
Billing is hourly. For 1 million vectors at 1536 dimensions, Qdrant Cloud runs $25-45/month -- comparable to Pinecone Serverless at $25-60/month for similar data.
Architecture
Qdrant's Rust core provides predictable latency under load -- the garbage-collection pauses that affect Java/Go-based vector stores are absent. It supports payload filtering (filter by metadata at query time), named vectors (multiple vector spaces per object), and sparse + dense hybrid search.
Horizontal scaling is handled via distributed mode: shards across nodes, with configurable replication.
Where it fits
Qdrant is the self-hosting choice. Teams with data residency requirements, teams that want to avoid SaaS vendor lock-in, and teams that need to run vector search in air-gapped environments all reach for Qdrant first.
The trade-off against Pinecone: Qdrant requires operational ownership (sizing, patching, backups) when self-hosted. Qdrant Cloud removes that, but at that point the cost comparison against Pinecone Serverless is roughly even and the managed experience is similar.
Native integrations with LangChain and Mastra mean drop-in use in most agent frameworks.
Field notes
- Qdrant Cloud standard cluster billing confirmed at $0.078/GB-hour in the 2026 pricing page. At 8GB RAM (a common sizing for 5-10M vectors), that is $45/month -- competitive with Pinecone Serverless for the same data volume. [changelog, 2026-04-01]
See also
Field notes synthesized from build evidence ; postmortems, dev-team blogs, and vendor retros. Methodology is public. Corrections to hello@vybing.dev.