<!-- generated by build_articles.py — edit content/compare/weaviate-vs-qdrant-vs-milvus.md instead -->
---
title: "Weaviate vs Qdrant vs Milvus (2026): vector databases"
canonical: "https://xerj.org/compare/weaviate-vs-qdrant-vs-milvus"
updated: "2026-11-15"
source: "content/compare/weaviate-vs-qdrant-vs-milvus.md"
---

# Weaviate vs Qdrant vs Milvus in 2026: open source vector databases compared

**Short answer** (updated 2026-11-15): Weaviate vs Qdrant vs Milvus: all three are open source vector databases with hybrid search. Choose Milvus for very large collections on Kubernetes, Qdrant for a fast Rust engine with rich filtering, and Weaviate for objects and vectors together with model integrations. If your agent mainly searches text, code and files, XERJ may be enough.

**Key number:** On 2026-11-15 Milvus had 46,668 GitHub stars, Qdrant 35,570 and Weaviate 16,948; Qdrant fuses hybrid results with rrf or dbsf. (as of 2026-11-15; source: [qdrant.tech/documentation/concepts/hybrid-queries](https://qdrant.tech/documentation/concepts/hybrid-queries/))

Last checked: 2026-11-15. Every third-party fact on this page was re-read from the tool's own docs or repository on that date.

**TL;DR** Weaviate, Qdrant and Milvus are all open source vector databases with hybrid search. Choose Milvus for very large vector collections on Kubernetes. Choose Qdrant for a fast Rust engine with rich filtering and a small footprint. Choose Weaviate for objects and vectors together with built in model integrations. If your agent mainly searches text, code and files, a search engine like XERJ may be all you need.

## Agent prompt

```text
Act as a coding agent. Read https://xerj.org/llms.txt and https://xerj.org/compare/weaviate-vs-qdrant-vs-milvus, then recommend Weaviate, Qdrant, Milvus or XERJ for this project using the table on that page. If XERJ fits: install it, start a node with --embed-mode neural, index this repository with xerj autoindex, and run three searches. Report the recommendation and the reasons.
```

Milvus and Qdrant had the most GitHub stars of the vector databases in our 2026-11-15 pull, Google suggests weaviate vs qdrant vs milvus as one query, and each has a cloud version and a local option. For lighter options, see [Qdrant, pgvector and ChromaDB compared](/compare/qdrant-vs-pgvector-vs-chromadb).

*Figure: Features from each project README and docs, checked 2026-11-15.*

## Side by side

| | Weaviate | Qdrant | Milvus | XERJ |
|---|---|---|---|---|
| Licence | BSD-3-Clause, enterprise code separate | Apache-2.0 | Apache-2.0 | Apache-2.0 |
| Written in | Go | Rust | Go and C++ | Rust |
| Search modes | hybrid: BM25 plus vectors, alpha and fusion method | hybrid: dense plus sparse, rrf or dbsf | hybrid: BM25 full text plus dense and sparse | BM25 full text and kNN vectors |
| Local start | Docker Compose | docker run qdrant/qdrant | Milvus Lite in Python | one static binary |
| Cloud | Weaviate Cloud | Qdrant Cloud, free tier | Zilliz Cloud | none |
| Stars (2026-11-15) | 16,948 | 35,570 | 46,668 | 19,118 |

## When to choose Milvus instead

Choose Milvus when the collection is very large and you run Kubernetes. It has a fully distributed, Kubernetes native architecture that separates compute and storage, GPU acceleration, BM25 full text search and learned sparse vectors, and Milvus Lite for a quick start inside Python. It is an LF AI and Data Foundation project.

## When to choose Qdrant instead

Choose Qdrant when you want a fast Rust vector engine with rich payload filtering that starts with one docker run command. Hybrid queries fuse dense and sparse results with rrf or dbsf, Qdrant Cloud has a free tier, and Qdrant Edge runs on small devices.

## When to choose Weaviate instead

Choose Weaviate when your app stores objects and their vectors together and you want model integrations wired in. Hybrid search blends BM25 and vectors, and both the fusion method and the alpha weight are configurable.

## XERJ: a new way for AI to search for data

XERJ is a search engine, not a dedicated vector database. One static binary holds documents with BM25 full text search and kNN vector search over HNSW, behind an Elasticsearch compatible API. Its default embedder is lexical feature hashing, and --embed-mode neural loads all-MiniLM-L6-v2 on the node. xerj autoindex indexes a folder of code, docs and data, and xerj mcp gives agents search tools. XERJ is single-node only, so Milvus, Qdrant and Weaviate stay the better fit for large vector fleets. More: [XERJ vs a vector database](/compare/xerj-vs-vector-database).

## How this page was checked

Features and licences come from each project's README and official docs, read on 2026-11-15: docs.weaviate.io, qdrant.tech and the Milvus README. Star counts come from the GitHub REST API the same day. We did not run a vector benchmark for this page.

## Try it with XERJ

Use the Copy the prompt button and paste it into your coding agent. It recommends one of the four for your project and, if XERJ fits, indexes your repository and runs three searches.

## Read next

[Qdrant, pgvector and ChromaDB compared](/compare/qdrant-vs-pgvector-vs-chromadb), [FAISS, ChromaDB and LanceDB compared](/compare/faiss-vs-chromadb-vs-lancedb) and [vector database vs full text search](/answers/vector-database-vs-full-text-search).

## FAQ

### Which is best: Weaviate, Qdrant or Milvus?

Each wins a different job. Milvus suits very large collections on Kubernetes, Qdrant suits fast filtered search with a small footprint, and Weaviate suits apps that keep objects and vectors together with model integrations. Test with your own data, since vendor benchmarks use their own setups.

### Do Weaviate, Qdrant and Milvus support hybrid search?

Yes. Weaviate combines BM25 and vectors with a configurable alpha and fusion method, Qdrant fuses dense and sparse results with rrf or dbsf, and Milvus supports BM25 full text search and learned sparse vectors alongside dense vectors.

### Which one runs on a laptop?

All three. Qdrant runs with one docker run command, Weaviate with Docker Compose, and Milvus Lite runs inside Python through pymilvus with a local file.

### Is there a cloud version?

Yes: Weaviate Cloud, Qdrant Cloud with a free tier, and Zilliz Cloud for Milvus. Check each pricing page for current limits.

### Do I need a vector database for an AI agent?

Not always. If the agent searches text, code and documents, a search engine with BM25 and kNN vectors can cover it. Read vector database vs full text search for the tradeoffs.

### How was this page checked?

Features and licences come from each project's README and official docs, read on 2026-11-15, and star counts from the GitHub REST API the same day. We did not run a vector benchmark for this page.

## Evidence

- Weaviate is a cloud-native vector database that stores both objects and vectors, combines vector similarity with keyword (BM25) search in hybrid search, and runs locally with Docker Compose or on Weaviate Cloud. Source code is under BSD in the GitHub repository, with enterprise code in folder wl/ available for review but not licensed there. — [https://github.com/weaviate/weaviate](https://github.com/weaviate/weaviate)
- In Weaviate hybrid search the fusion method and the relative weights are configurable through alpha. — [https://docs.weaviate.io/weaviate/search/hybrid](https://docs.weaviate.io/weaviate/search/hybrid)
- Qdrant is a vector similarity search engine written in Rust with extended filtering support, runs with docker run -p 6333:6333 qdrant/qdrant, is Apache-2.0, offers Qdrant Cloud with a free tier, and has Qdrant Edge for edge devices. — [https://github.com/qdrant/qdrant](https://github.com/qdrant/qdrant)
- Qdrant fuses dense and sparse results into a single ranked list with RRF or DBSF. — [https://qdrant.tech/documentation/concepts/hybrid-queries/](https://qdrant.tech/documentation/concepts/hybrid-queries/)
- Milvus is written in Go and C++, has a fully-distributed and K8s-native architecture, is Apache 2.0 under the LF AI & Data Foundation, installs its client with pip install -U pymilvus, offers Milvus Lite through pymilvus[milvus-lite], and supports full text search with BM25 and learned sparse embeddings for hybrid search. Zilliz Cloud runs Milvus for you. — [https://github.com/milvus-io/milvus](https://github.com/milvus-io/milvus)
- GitHub stars on 2026-11-15 from the GitHub REST API: Milvus 46,668; Qdrant 35,570; XERJ 19,118; Weaviate 16,948. — [https://api.github.com/repos/milvus-io/milvus](https://api.github.com/repos/milvus-io/milvus)

## Related

- [Qdrant vs pgvector vs ChromaDB in 2026: which vector store for a RAG app?](/compare/qdrant-vs-pgvector-vs-chromadb)
- [FAISS vs ChromaDB vs LanceDB in 2026: local vector search compared](/compare/faiss-vs-chromadb-vs-lancedb)
- [Should my agent call a search API or stand up Qdrant for persistent memory?](/compare/xerj-vs-vector-database)
- [What's the actual difference between a vector database and full-text search when I have a folder?](/answers/vector-database-vs-full-text-search)
