Milvus

Open source vector database for similarity search at scale, also available via Zilliz Cloud

Data & analytics

Overview

Milvus is an open source vector database licensed under Apache 2.0, designed for similarity search at scale with native support for sharding and replication across billions of vectors. The software is free and self-hosted via Docker or Kubernetes. Zilliz, the company that created Milvus, offers a managed version called Zilliz Cloud: a Free plan at $0 with 5 GB of storage and 2.5 million vCU included per month, a Standard plan starting at $0/month serverless (usage-based billing) or from $126/GB/month for a dedicated cluster, an Enterprise plan starting at $197/month with 99.95% SLA, SSO, and private VPC, and a Business Critical plan on request for regulated sectors.

Milvus is not translated into French, neither on the project website nor on the Zilliz Cloud console. Zilliz documents multi-region hosting including Europe on AWS, GCP, and Azure for dedicated clusters, which allows you to keep data in the EU zone. The API relies on Python, Java, Go, and Node.js SDKs, with official integrations to LangChain, LlamaIndex, Haystack, Dify, Langflow, and n8n. The main limitation is the complexity of self-hosting: Milvus requires a distributed architecture (etcd, object storage, message queue) that few small teams can maintain alone, which almost inevitably pushes toward the managed Zilliz Cloud offering once you move beyond prototyping.

Our verdict

Best for projects anticipating billions of vectors that need a proven distributed architecture, either self-hosted or via Zilliz Cloud. Not for you for a first prototype or a small team without systems administration skills: the complexity of self-hosting Milvus is significantly higher than Chroma or Qdrant.

← Back to all tools
Milvus: pricing, review and alternatives — librairy.io