Qdrant
High-Performance Vector Search at Scale.
GitHub Stars
30k+
Community Members
60k+
Core Language
Rust
Compliance
SOC2 & HIPAA
About Qdrant
Qdrant is a powerful vector search engine engineered for real-time retrieval at any scale. It offers a rich feature set for modern AI applications, including advanced JSON metadata filtering (nested, geo, text) and native hybrid search blending keyword (BM25, SPLADE++) and vector search in a single query. Its one-stage filtering mechanism applies filters during HNSW traversal for high recall and low latency. Qdrant also supports multivector capabilities for more expressive and multimodal retrieval, along with built-in reranking logic like score boosting and Maximum Marginal Relevance (MMR) to refine search results. The engine can be deployed via a managed cloud, hybrid cloud, or self-hosted for maximum flexibility.
Search & Filtering Capabilities
Hybrid Search
Natively blends keyword (sparse) and vector (dense) search in one query. Supports BM25, SPLADE++, and miniCOIL.
Metadata Filtering
Store metadata in JSON and use advanced filters, such as nested, text, geo, and has_vector.
One-Stage Filtering
Filters are applied during the HNSW graph traversal, ensuring high recall with low latency without pre- or post-filtering.
Reranking
Infuse business logic with score boosting, use late interaction models like ColBERT, and diversify results with Maximum Marginal Relevance (MMR).
Multivector Support
Allows storing multiple vectors per object for more expressive and multimodal retrieval.
Deployment Models
Qdrant Cloud
Fully managed, serverless vector database with a free tier and pay-as-you-go pricing.
Open Source
Self-host and manage Qdrant on your own infrastructure using Docker or Kubernetes.
Enterprise
Provides dedicated clusters, expert support, and custom features for large-scale deployments.
Hybrid Cloud
Deploy Qdrant in your own VPC for data privacy while retaining a managed control plane.