Qdrant

Qdrant develops a vector similarity search engine and vector database written in Rust, designed to store, search and manage "points" — vectors with attached JSON payloads. The engine supports dense vectors for semantic similarity, sparse vectors for lexical/keyword matching (including BM25, SPLADE++ and miniCOIL), and multivector/late-interaction models such as ColBERT. Distinguishing capabilities described by the company include one-stage filtering applied during HNSW graph traversal via payload indexes (rather than pre- or post-filtering), configurable hybrid fusion strategies such as Reciprocal Rank Fusion and Distribution-Based Score Fusion, reranking through score boosting and Maximum Marginal Relevance, scalar/product/binary and asymmetric quantization, on-disk and memory-mapped storage, io_uring optimizations, multitenancy, and horizontal scaling through sharding and replication with zero-downtime updates.

The product is distributed as an open-source server (Docker image, REST/OpenAPI 3.0 and gRPC interfaces, official Python, JavaScript/TypeScript, Rust, Go, .NET/C# and Java clients, plus community Kotlin and PHP clients) and as commercial deployment options: Qdrant Cloud (fully managed on AWS, GCP and Azure with high availability and auto-sharding), Qdrant Hybrid Cloud (customer-run Kubernetes with decoupled control and data planes), Qdrant Private Cloud (air-gapped on-premises), and Qdrant Edge, a beta in-process library for edge and resource-constrained devices that syncs with a Qdrant server. Additional platform features include a built-in web UI for exploring collections, Cloud Inference for server-side text and image embedding generation, monitoring integrations (Prometheus, Grafana, Datadog), SSO via SAML/OIDC, RBAC, private networking, backups with point-in-time restore and vector-scoped API keys. The company states SOC 2 and HIPAA compliance and GDPR-aligned options.

Typical use cases the company targets are retrieval-augmented generation and generative AI, AI agents with persistent memory, semantic search, recommendation systems, and anomaly/outlier detection.

Founding story

In 2021 André Zayarni and Andrey Vasnetsov worked together on a project to build a matching engine for unstructured data objects using vector similarity search. After evaluating existing options, including the FAISS library, they concluded that none met their feature and scalability requirements, so Vasnetsov built his own production-ready vector search engine from scratch. The first version was published on GitHub and attracted developer interest, and the feedback from developers and startups led the founders to formally establish Qdrant as a company.

Business model

Open-core: the core vector search engine is published as open source on GitHub, while revenue-generating offerings are managed and self-hosted commercial products — Qdrant Cloud (managed service with a free tier), Hybrid Cloud, and Private Cloud — plus enterprise support. Managed and hybrid tiers add high availability, zero-downtime upgrades, monitoring and alerting, a central management UI, automated backups and disaster recovery, and enterprise support that are not part of the open-source distribution.

Cloud subscription and enterprise plans layered on a free open-source engine; the company advertises a free tier in Qdrant Cloud, managed cloud pricing, and enterprise plans for Hybrid and Private Cloud deployments. Specific pricing figures are not disclosed in the sources.

Traction

The open-source project reports more than 250 million package downloads and 29,000+ GitHub stars as of the March 2026 Series B announcement, with the GitHub repository showing 34.1k stars and 2.6k forks and 6,679 commits at the time of collection. The company site cites 30k+ GitHub stars and 60k+ community members, while the about-us page cites 29K stars and 9k community members. Named production deployments include Canva, Tripadvisor (multimodal retrieval across billions of signals for an AI trip planner), HubSpot, OpenTable, Bazaarvoice, Roche, Bosch, Dust (vector search across 5,000+ data sources) and Lyzr (reported 90% latency reduction and 150% throughput increase). Another customer case cites real-time context across 2M+ AI-driven conversations.

Latest developments

On 12 March 2026 Qdrant announced a $50 million Series B led by AVP with Bosch Ventures, Unusual Ventures, Spark Capital and 42CAP participating, positioning the company around "composable vector search" as core AI infrastructure. Alongside the raise, the company highlighted Qdrant Edge for resource-constrained devices with cloud sync, agent skills for AI coding assistants, Cloud Inference for server-side embedding generation, and community contributions in recent releases (ASCII folding for multilingual full-text retrieval in v1.16 and field-level control over HNSW graph construction). The company also states it is hiring.