Qdrant

Qdrant is an open source vector search engine and database for building AI retrieval: semantic search, retrieval augmented generation, recommendation systems, agent memory and anomaly detection. Written in Rust with a custom storage engine, it stores vectors alongside JSON metadata, indexes new data in real time without rebuilding the index, and applies metadata filters during HNSW traversal rather than before or after the search, which keeps recall high under complex conditions. Scalar, binary and asymmetric quantization reduce memory use for very large collections. It serves engineering teams building retrieval systems in any deployment model. Native hybrid search blends dense and sparse vectors in one query, with support for BM25, SPLADE and miniCOIL. Multivector storage supports late interaction models such as ColBERT, and reranking options include score boosting and maximum marginal relevance. Developers work through REST or gRPC and official clients for Python, JavaScript and other languages, with a built in web UI for exploring collections. Qdrant Cloud adds managed clusters, cloud inference that generates text and image embeddings inside the service, multitenancy, role based access control, private networking, backups with point in time restore and single sign on. Hybrid Cloud runs managed clusters inside a customer's own infrastructure, and Private Cloud offers fully isolated or air gapped deployments. The engine itself is free to self host. Qdrant Cloud pricing is published in tiers: a Free tier with a single node cluster of 0.5 vCPU, 1 GB RAM and 4 GB disk; a Standard tier with usage based pricing, dedicated resources, backups and a 99.5% uptime SLA; and a Premium tier with a minimum spend that adds SSO, private VPC links, a 99.9% SLA and extra support. Hybrid Cloud and Private Cloud are quoted by the sales team, and a pricing calculator estimates cluster costs.