
The open-source (Apache 2.0) vector database built for billion-scale search, with Zilliz Cloud as its managed service — free to self-host, or serverless and dedicated clusters with a free tier.
Milvus is the open-source vector database most often shortlisted against Pinecone when a team expects very large collections or wants to self-host. It ships in three shapes — Milvus Lite as a pip-installed library for prototyping, Standalone on one machine, and Distributed for horizontal scale to billions of vectors — under the Apache 2.0 license, and Milvus 3.0, released in July 2026, moves storage to a lake-native, S3-based design. The managed path is Zilliz Cloud, from the company that builds Milvus: a free tier with 5 GB of storage and 2.5 million vCUs a month, a Standard plan from $0 on serverless or $126 a month on dedicated, Enterprise from $197 a month on dedicated, and Business Critical and BYOC by quote. Zilliz names customers including NVIDIA, Reddit, Airtable, and Filevine; vendor-reported. The flexibility has a cost: self-hosted Distributed is a real cluster to run, and Zilliz pricing mixes vCUs, plan minimums, and per-million-vector rates that take modelling. Buy Milvus if you need billion-scale vectors, want open source you can self-host, or want a managed option with BYOC. Buy Pinecone if you want the least operational work, or Chroma or Qdrant for a lighter footprint at small to mid scale.
Get a new AI workflow each week — many feature Milvus and other tools in this category.