Qdrant

Qdrant is an open-source vector search engine and database purpose-built for the demands of production AI applications. Written entirely in Rust with SIMD optimization and a custom storage engine called Gridstore, Qdrant is designed to deliver fast, memory-efficient similarity search at scale, going beyond the static-benchmark limitations of early vector libraries by offering persistence, payload storage, real-time indexing, and full production-grade operational tooling. Its retrieval layer combines dense and sparse vector search in a single query (supporting BM25, SPLADE++, and miniCOIL), applies metadata filters directly during HNSW graph traversal for high recall at low latency, and supports built-in multivector representations, score boosting, and late interaction reranking models like ColBERT.

Founded in 2021 in Berlin by CEO Andre Zayarni and CTO Andrey Vasnetsov, Qdrant grew out of the founders' work building a matching engine for large volumes of unstructured text data, pivoting toward neural search and open-sourcing the engine on GitHub the same year. The company has since raised a $28 million Series A led by Spark Capital and a $50 million Series B in March 2026 led by AVP with participation from Bosch Ventures, Unusual Ventures, Spark Capital, and 42CAP, positioning itself around the idea that "retrieval is what makes models useful in production." Qdrant is used across RAG pipelines, AI agent memory, recommendation systems, semantic search, and anomaly detection, with production case studies from Tripadvisor, HubSpot, Deutsche Telekom, Dust, and Canva.

Qdrant can be deployed however a team needs: as a fully managed Qdrant Cloud cluster on AWS, GCP, or Azure; as Hybrid Cloud running on a customer's own Kubernetes infrastructure while still managed through Qdrant's control plane; as an air-gapped Private Cloud for the strictest compliance needs; or as lightweight Qdrant Edge for low-latency search close to where data is generated. The platform is SOC 2 and HIPAA compliant with GDPR-aligned options, and offers developer-friendly REST and gRPC APIs, official client libraries, a built-in Web UI for visual exploration, and native Cloud Inference for generating embeddings without standing up a separate pipeline.

Qdrant's Free Tier is free forever, offering a single-node cluster with 0.5 vCPU, 1GB RAM, and 4GB disk, plus free cloud inference with selected models, ideal for testing and prototypes. The Standard Tier uses usage-based pricing for production workloads, billed hourly based on compute (vCPU), memory (GB), storage (GB), backup storage, and used inference tokens for paid models; it includes dedicated resources, flexible vertical and horizontal scaling, high availability setups, backup and disaster recovery, and a 99.5% uptime SLA. The Premium Tier requires a minimum spend and adds SSO, private VPC links, a 99.9% uptime SLA, and extra support for enterprises with additional security and compliance needs, available by contacting sales. Qdrant Hybrid Cloud lets teams run managed Qdrant clusters on their own infrastructure for local data residency and regulated workloads, while Private Cloud offers a fully isolated, air-gapped deployment for large enterprises; both require contacting the Qdrant team for pricing. The open-source Qdrant engine itself remains free and self-hostable under an Apache 2.0 license.