Qdrant

What it does

Qdrant is a vector similarity search engine and vector database. You store embeddings as points, each a vector with an attached JSON payload, then search them with filters. It accepts dense, sparse and multivector representations and returns ranked matches. Developers use it to build retrieval for RAG, semantic search, recommendations and anomaly detection, and it can rerank candidates with late interaction models such as ColBERT.

Work starts with a single Docker command or a Qdrant Cloud cluster, then a client library or the REST or gRPC API. Its distinguishing features are filtering applied during HNSW traversal, hybrid dense and sparse fusion, and quantization that can reduce RAM use by up to 97%. A built-in web UI lets you inspect collections and test queries.

Qdrant suits engineering teams building AI retrieval who can run or manage a database service. The core is open source under Apache 2.0 and the self-hosted route needs no account. Qdrant Cloud has a free tier of 1GB RAM and 4GB disk for testing and prototypes without high availability. Qdrant Edge is labelled Beta, and paid Cloud tiers are usage-based, so check the pricing calculator for exact costs.

What it can do

• Dense, sparse and multivector search over stored embeddings
• Filtering on JSON payload with nested, text, geo and numeric conditions
• Hybrid search that fuses dense and sparse results using RRF or DBSF
• Reranking with Maximal Marginal Relevance and late interaction models such as ColBERT
• Vector quantization and on-disk storage to reduce RAM use
• Sharding and replication for horizontal scaling with zero-downtime updates
• Snapshots, backups and point-in-time restore on Qdrant Cloud

Example workflows

Hybrid retrieval with ColBERT reranking

Input: documents encoded as dense vectors and ColBERT multivectors via FastEmbed. Steps: create a collection with HNSW disabled on the multivector field, upload points with batch_size, then run query_points with a dense prefetch and a ColBERT rerank. Output: the top-ranked matching documents with their payload text.

Filtered semantic search over payloads

Input: a text query embedded as a dense vector plus payload conditions such as category or numeric range. Steps: attach JSON metadata when upserting points, then query with must, should or must_not filters. Output: nearest neighbours that also satisfy the filter conditions.

Before you choose

• The Docker quickstart starts an insecure deployment without authentication on all network interfaces, so secure it before production use.
• Disabling HNSW on multivector fields saves memory and speeds inserts but means those fields serve reranking rather than first-pass retrieval.
• The free cloud tier is limited to 1GB RAM and 4GB disk without high availability, so it is for testing and prototypes.

Listed under

• Coding
• Sandboxes and infrastructure
• AI agents
• Database
• Developers
• Hybrid search
• Managed cloud
• Open source

Get started

Sign up for a free Qdrant Cloud cluster, or run docker run -p 6333:6333 qdrant/qdrant for a local instance; secure it before exposing it to a network.

1. Run the container with docker run -p 6333:6333 qdrant/qdrant, or sign up at cloud.qdrant.io for a managed cluster.
2. Install a client library such as the Python client and connect with QdrantClient(url="http://localhost:6333").
3. Read the installation and security guides, then enable authentication before exposing the instance to a network.

Qdrant Cloud is billed on actual resource usage (vCPU, memory, storage, backups and paid inference tokens), charged hourly. The free tier is 1GB RAM and 4GB disk. Use the Pricing Calculator for estimates.

Source: qdrant.tech/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Qdrant - Vector Search Engine ↗qdrant.tech
• Pricing for Cloud and Vector Database Solutions Qdrant ↗qdrant.tech
• GitHub - qdrant/qdrant: Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/ · GitHub ↗github.com
• Multivectors and Late Interaction - Qdrant ↗qdrant.tech

Company website

qdrant.tech: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.