Qdrant
High-performance open-source vector database written in Rust
Our Take
Qdrant is the performance-focused choice with the simplest operational profile among purpose-built vector DBs — a single Rust binary with zero external dependencies. Its standout feature is integrated HNSW filtering: filters are applied during graph traversal (not before or after), making it best-in-class for workloads with heavy metadata filtering. Supports scalar, product, and binary quantization (32x memory reduction). Flat latency curve regardless of k (22–24ms from k=1 to k=100, vs Weaviate's linear drift). ~30,000 GitHub stars, 25M+ Docker downloads. Cloud free tier gives 1GB RAM forever. The tradeoff: HNSW-only (no IVF or DiskANN), ingestion can degrade query latency on the same node, and less proven at multi-billion scale than Milvus. $50M Series B raised March 2026.
Pros
- + Excellent performance (Rust, single binary, zero deps)
- + Best-in-class filtered search (filters during HNSW traversal)
- + Binary quantization for 32x memory reduction
- + Built-in multi-tenancy
- + Low operational complexity
- + Embedded mode available for prototyping
Cons
- - HNSW-only (no IVF or DiskANN alternatives)
- - Ingestion can degrade query latency on same node
- - Less proven at multi-billion scale
- - No built-in vectorization
Details
- Pricing Model
- freemium
- Starting Price
- $0
- Self-Hosted
- Yes
- Cloud Hosted
- Yes
- Founded
- 2021
- Repository
- GitHub →
Best For
- • High-performance filtered vector search
- • Single-binary simplicity (no external deps)
- • Multi-tenant applications
- • Teams wanting open source with low ops overhead
Integrations
Articles featuring Qdrant
Last updated: