Qdrant
Qdrant is an open-source vector database and similarity search engine that stores “points” (vectors plus payload). The site says it is built entirely in Rust with SIMD and a custom storage engine called Gridstore.
Repository snapshot (GitHub API, 2026-11-19)
| Item | Detail |
|---|---|
| Repo | qdrant/qdrant, Apache-2.0 |
| Created | 2020-05-30 |
| Activity | last push 2026-11-19; latest release v1.20.2 (2026-11-16) |
| Size | 35,637 stars, 2,825 forks |
| Maintainer | Qdrant (company) |
Capabilities (qdrant.tech)
- Hybrid search of dense and sparse vectors, naming BM25, SPLADE++ and miniCOIL; see hybrid-search-and-rank-fusion.
- Payload filtering (nested, text, geo) applied during HNSW traversal.
- Quantisation: asymmetric, scalar and binary, with up to 64x memory reduction claimed.
- Multitenancy, role-based access control and vector-scoped API keys.
- Reranking, including ColBERT-style late interaction and MMR; real-time indexing, point-in-time backups and multivector support.
Deployment and billing
Self-hosted (Docker, on-prem), Hybrid Cloud (your own Kubernetes), Private Cloud (air-gapped), Edge (beta) and the managed Qdrant Cloud on AWS, GCP or Azure. Plan names: Free, Standard (usage-based) and Premium (minimum spend). A serverless option is on a waitlist. Rates are on the vendor’s pricing page and are not recorded here. The site states no embedded in-process mode and does not list index types beyond HNSW.
Sources
- qdrant.tech, pricing, the Qdrant README and GitHub API (2026-11-19)