Qdrant Vector Search Engine

Get Filter-Aware Vector Indexing on Qdrant Cloud

Qdrant indexes your vectors with HNSW and your metadata with payload indexes, so filtered search stays accurate and fast. Start free with no credit card.

HNSW index

Tune One HNSW Index for Speed and Recall

Qdrant indexes dense vectors with HNSW, the graph behind fast similarity search at scale. It finds nearest neighbors in a fraction of the comparisons a full scan takes.

A payload index on a field adds filter-aware links between graph points that share a value.
  • Tune build and search separately.

    Set m and ef_construct when the graph is built, and raise hnsw_ef per query for higher recall.

  • Check recall against exact results.

    Set exact: true on a query to run a full scan and compare the approximate results with it.

  • Rebuild faster, on GPUs too.

    Qdrant reuses the existing HNSW index to speed up the next build, and Qdrant Cloud supports GPUs for indexing on AWS.

Start free with Qdrant Cloud

Payload indexes

Index Your Metadata Next to Your Vectors

Qdrant combines its vector index with payload indexes on the fields you filter on, so the vector index reads payload values during the search.

Get filter-aware links.

When Qdrant builds the HNSW graph, it adds extra edges for fields that already have payload indexes, which keeps filtered search accurate and fast.

Pick an index type per field.

Index keyword, integer, float, datetime, geo, and full-text fields, with a stemmer for your language on text.

Index sparse vectors too.

Sparse vectors get their own inverted index, built in the same optimization process as the dense index.

LLM infrastructure

Fit Qdrant Into Your LLM Infrastructure

In LLM application architecture, Qdrant is the retrieval layer. It finds the context each prompt needs, which puts LLM context management inside the query.

Scope each prompt's context.

Filter on user, product, or date range during the search, so every passage already matches the request.

Get memory-efficient storage.

Scalar quantization stores vectors in 4x less memory than float32, and binary or product quantization compresses them further to cut cost.

Keep your framework.

Qdrant integrates with more than 40 AI agent and RAG frameworks, including LangChain, LlamaIndex, and Haystack.

Start free with Qdrant Cloud

Results

Indexes Serving Production Search on Qdrant

Sapu

28 million PubMed abstracts in a single collection, embedded at 3,072 dimensions, on Qdrant Cloud.

Read the Sapu case study

Bazaarvoice

about 98% nearest neighbor recall with sub-100 ms queries over billions of vectors on Qdrant Hybrid Cloud.

Read the Bazaarvoice case study

Himalayas

about 30 ms median and 200 ms p95 inside Qdrant across 25M+ requests a month.

Read the Himalayas case study
Start free with Qdrant Cloud

Getting started

From Sign-Up to Your First Indexed Collection

Create a free cluster.

Sign up for Qdrant Cloud with no credit card.

Set up the collection.

Choose the vector size and distance metric, and add the payload fields you plan to filter on.

Upload and query.

Upload your vectors with their payloads, then search with a vector and a filter.

Start Free

Common Questions

How do I get started with Qdrant Cloud?

A free Qdrant Cloud cluster has 1 GB RAM, 0.5 vCPU, 4 GB disk, and one node, enough for about 1 M vectors of 768 dimensions, with no credit card. Move to a Standard cluster when you need dedicated resources, multi-node high availability, backups, and SLAs.

Does Qdrant use an IVF index?

Qdrant, an open source engine written in Rust, uses HNSW as its dense vector index, so you tune one graph with m, ef_construct, and hnsw_ef, and payload indexes add filter-aware links to it.

When should I create payload indexes?

Before you upload data. Qdrant adds filter-aware links for the payload indexes that exist when the graph is built. A payload index added later needs an HNSW rebuild.

Index Your Vectors on Qdrant Cloud

Tune HNSW for your recall target and keep filtered queries fast as your data grows.

Rocket flying over globe illustration