Milvus (Open Source)
Free
- Apache 2.0 license
- Self-hosted on Kubernetes
- Full feature access
- Community support
Milvus — Vector Databases Platform
Founded 2017|San Francisco, California|51-200 people|Open Source
What is Milvus?
Milvus is an open-source vector database built for scalable similarity search, capable of handling billions of vectors in distributed environments. Created by Zilliz, a company founded in 2017 by Charles Xie (former founding engineer of Oracle 12c cloud database), Milvus has become one of the most widely deployed vector databases with over 30,000 GitHub stars.
The database supports multiple index types including IVF, HNSW, and DiskANN, with GPU-accelerated search and hybrid search combining dense and sparse vectors in a single query. Milvus runs on Kubernetes for production deployments and is governed under the LF AI & Data Foundation. Zilliz Cloud offers a fully managed version with automatic scaling, starting with a free tier and usage-based pricing from $4 per million vector compute units.
Zilliz has raised approximately $113-132 million in funding, with a $60 million Series B extension in August 2022 led by Prosperity7 Ventures (Aramco). The company is headquartered in San Francisco with roughly 140 employees. Zilliz was named "Highest Performer" and "Easiest to Use" in G2's Summer 2025 Vector Database Grid Report.
Key features
Core capabilities this platform advertises.
- Billion-scale vector search
- GPU-accelerated indexing
- Distributed architecture
- Multiple index types
- Cloud and self-hosted options
Strengths and tradeoffs
What this tool does well, and the limitations to keep in mind.
Pros
- Extreme scalability handling billions of vectors in distributed environments
- Fully open-source under Apache 2.0 with no vendor lock-in
- Hybrid search combining dense and sparse vectors in single queries
- Rich index flexibility with IVF, HNSW, DiskANN and GPU acceleration
- Managed cloud option reduces DevOps burden with auto-scaling
Cons
- Complex distributed architecture requiring Kubernetes for production
- Steep learning curve for index tuning and deployment configuration
- High memory and storage requirements for large datasets
- No in-place vector updates requiring delete and re-insert workarounds
Plans & pricing
What's included in each plan, and how the tiers compare.
Zilliz Cloud Free
$0
- 5GB storage
- 2.5M vCUs/month
- Up to 5 collections
Zilliz Cloud Standard
From $99/month
- Up to 32 query CUs
- Manual scaling
- Email/ticket support
Zilliz Cloud Enterprise
From $155/month
- 99.95% uptime SLA
- SSO and RBAC
- 24/7 enterprise support
- VPC peering
Common use cases
Organizations that need vector search at billion-scale with high throughput
- Billion-vector similarity search
- Large-scale recommendation systems
- Image and video retrieval
- Genomics and scientific computing
- Enterprise-scale RAG systems
Using Milvus with Respan
Respan complements Milvus by providing observability for the LLM layer in RAG (Retrieval-Augmented Generation) pipelines. Teams using Milvus for vector search can monitor the end-to-end AI pipeline including embedding generation, retrieval quality, and LLM response generation through Respan.
- Monitor RAG pipeline performance from embedding to generation
- Track LLM costs in Milvus-powered retrieval workflows
- Evaluate retrieval quality and LLM response accuracy
- Optimize embedding model selection for vector search quality
Best Milvus alternatives & competitors
Top companies in Vector Databases you can use instead of Milvus.
Pinecone
Fully managed serverless vector database
Qdrant
High-performance open-source vector search
Chroma
Lightweight embedded vector database
Supabase
pgvector hosting
Weaviate
Open-source vector database
Neo4j
Native graph database with Cypher query language
MongoDB Atlas Vector Search
Elasticsearch
Redis Vector
ClickHouse
Analytics + vector
TigerGraph
Vespa
Neon
Serverless Postgres
LanceDB
Turbopuffer
SingleStore
ArangoDB
Compare Milvus
Side-by-side comparisons with other tools in this category.
Best integrations for Milvus
Companies from adjacent layers in the AI stack that work well with Milvus.
Last verified: February 28, 2026