Tools/milvus-io/milvus

Milvus

Cloud-native vector database for scalable ANN search

46.1k ★+18/wkgrowthGoApache License 2.0 →trending

View on GitHub →Compare in Vector Databases →Homepage →

Replaces:Pinecone

See how it compares →

The Lens

By Erik Loyd, SaaS CEO and former COO/CFO of an AWS Premier Partner.

Updated Aug 2026

When your app converts text or images into numerical vectors (via OpenAI, Cohere, or any embedding model), Milvus finds the closest matches across millions or billions of vectors in milliseconds. Apache 2.0, Go/C++. Cloud-native architecture with storage and compute separated. Supports multiple index types (HNSW, IVF, DiskANN), hybrid search (vectors + scalar filters), and multi-tenancy. The SDK supports Python, Java, Go, Node.js, and REST.

Self-hosting is free. Full feature set, no restrictions. Runs on Kubernetes with Helm charts or standalone via Docker. Minimum production setup: 3 nodes for high availability. Zilliz Cloud (managed Milvus) has a free tier with 2 collections and 1M vectors. Paid starts at ~$65/mo for a dedicated cluster. Serverless option available for variable workloads.

Solo: Milvus Lite (embedded mode) or Zilliz free tier for prototyping. Small teams: Zilliz free tier or self-host with Docker. Medium to large: self-host on Kubernetes or Zilliz Cloud depending on ops capacity.

The catch: self-hosted Milvus on Kubernetes is operationally heavy: etcd, MinIO, Pulsar (or Kafka) as dependencies. That's a lot of infrastructure for a vector search feature. If your dataset is under 1M vectors, Chroma or pgvector (Postgres extension) gets you there with dramatically less infrastructure.

Free vs Self-Hosted vs Paid

free self hosted paid cloud

Free (Self-Hosted)

Milvus is Apache 2.0 with nothing held back: every index type, hybrid search, and multi-tenancy. Milvus Lite runs in-process for local development and prototyping.

Zilliz Cloud

Free: $0. Prototyping scale with community support.
Serverless: from $0/mo, billed on usage. The right shape for spiky or low-volume workloads.
Dedicated: from $126/GB/mo for isolated, reserved compute measured in compute units.
Enterprise: from $197/mo on Dedicated, with enterprise support included.

Paid tiers carry a 30-day free trial. Zilliz also sells a bring-your-own-cloud deployment with the same features as their SaaS Dedicated clusters.

Self-Hosted Costs

Production Milvus is a Kubernetes deployment with etcd, object storage such as MinIO or S3, and a message queue. Three nodes minimum for high availability. Expect $100 to $500/mo of infrastructure depending on provider and data volume, plus a standing ops commitment. This is not a container you set up once and forget.

When to Pay

Pay for Zilliz when running Kubernetes for a vector database exceeds what your team wants to own, or when managed backups and scaling are worth the margin. Serverless is the honest starting point since it begins at $0 and scales with actual queries. Before either, check whether pgvector on the Postgres you already run is enough, because for a lot of workloads it is.

Self-hosting is free under Apache 2.0 but wants a Kubernetes cluster and $100 to $500/mo of infrastructure. Zilliz Cloud starts free, Serverless from $0/mo on usage, Dedicated from $126/GB/mo.

What to do by team size

Solo: free with Milvus Lite locally, or the Zilliz free tier for prototyping
Small team: free Docker self-host for small datasets, or Zilliz Serverless from $0/mo on usage
Medium team: Zilliz Dedicated from $126/GB/mo, or self-host if you already run Kubernetes
Large team: free self-hosted on your own Kubernetes, or Zilliz Enterprise from $197/mo with support

Self-hosting ops:heavy

Get tools like this every Wednesday

One featured tool, three on the radar. No fluff.

Similar Tools

Chroma88similar

Data infrastructure for AI

Qdrant88similar

High-performance vector database and search engine

Weaviate88similar

Open-source vector database

Score

88/100 · A

Adoption27/30

Maintenance25/25

Community11/20

License15/15

Analysis10/10

A low score is not a verdict on quality. Young and niche tools start low by design. How we calculate scores

Trust Signals

High adoption: 43,836 starsActive community: 3,962 forksPermissive license (Apache-2.0)

About

Stars: 46,134
Forks: 4,255

Explore Further

GitHub Repository

Source code, issues, README

Reddit Discussions

Community opinions and use cases

Hacker News

HN threads and discussions

Dev.to Articles

Tutorials and write-ups

Tutorials & Guides

Getting started resources

Official Website

Docs, blog, and more