Qdrant review: Open-source vector search engine with a free managed cloud tier

What Qdrant does

Stores vectors with JSON metadata and runs similarity and hybrid dense-sparse search for RAG, agents and recommendations, via API.

Qdrant is an open-source vector search engine written in Rust. It stores vectors together with JSON metadata and lets developers run similarity search through REST, gRPC or official clients such as Python and JavaScript. The vendor positions it for retrieval-augmented generation, AI agents, recommendations, semantic search and anomaly detection.

On the capability side, the official pages list hybrid search that mixes dense and sparse vectors, multiple vectors per object, metadata filters applied during HNSW traversal, and reranking options such as score boosting, late interaction models and Maximum Marginal Relevance. Quantization is offered to cut memory use, and new data is indexed without rebuilding the whole index. A built-in web UI lets you browse collections and test queries.

Qdrant can be used as the open-source engine, or through Qdrant Cloud, which runs on AWS, GCP or Azure. Hybrid Cloud and Private Cloud options cover running on your own infrastructure or in air-gapped setups.

Pricing for Qdrant Cloud has three tiers. The Free Tier is described as free forever, meant for testing and prototypes, with a single node cluster of 0.5 vCPU, 1GB RAM and 4 GB disk. The Standard Tier uses usage-based pricing for production workloads, billed hourly on compute, memory, storage, backups and paid inference tokens. The Premium Tier requires a minimum spend and adds SSO, private VPC links and a 99.9% uptime SLA. Pricing amounts are not published on the official pricing page.

• Open-source vector search engine in Rust
• Hybrid dense and sparse search
• JSON metadata filtering
• Multiple vectors per object
• Reranking with MMR and late interaction models
• Quantization to reduce memory
• Built-in web UI
• Cloud inference for embeddings

Our take

Qdrant fits developers and teams who need a dedicated vector database for RAG, search or recommendations and want the option to start on the open-source engine or a free cloud cluster. The Free Tier suits prototypes, but it is small and has no high availability, so production use means moving to the Standard Tier. The main trade-off is cost clarity: Standard is usage-based and Premium needs a minimum spend, with no amounts on the pricing page, so you need the pricing calculator or sales to estimate a budget. It is an infrastructure component, not an end-user app, so it is not for people who want a ready-made search interface without writing code.