Milvus

license

Apache-2.0

source

public

core-gated

ungated

Apache-2.0 and fully public, governed by LF AI & Data. The LICENSE file resolves to the full Apache-2.0 text, the repository carries the matching badge and docker-compose self-hosting instructions, and Zilliz Cloud is named only as the managed offering beside it rather than as a tier holding anything back, so nothing is withheld from the published tree.

• https://github.com/milvus-io/milvus/blob/master/LICENSE recorded 2026-08-13

  Full Apache License 2.0 text (11,357 bytes), served from raw.githubusercontent for the master branch.

• https://github.com/milvus-io/milvus recorded 2026-08-13

  Repo page - "Apache-2.0 license", 45,627 stargazers, docker-compose standalone install in the README, Zilliz Cloud referenced as the managed service rather than as a tier holding functionality back. No enterprise directory or separately licensed tree.

• https://en.wikipedia.org/wiki/Milvus_(vector_database) recorded 2026-08-13

  Infobox license Apache License 2.0; joined LF AI & Data January 2020 and "became a graduate in June 2021", citing the foundation's graduation announcement.

Banded on the downloads of the declared client SDK: where a server has no countable channel of its own, its client's downloads are the best available usage signal, and this product declares pymilvus on exactly that basis. The registry reports 5,976,463 downloads in the last month (1,484,328 in the last week, 241,978 in the last day), which lands in 1M-10M on the software scale. Qdrant is treated the same way, being the same shape of product and banded on qdrant-client. Milvus is the default open source engine for billion-scale semantic search and an LF AI & Data graduate, but the band rests on the count rather than on that description. Zilliz publishes no server-side deployment figure, and the repository's 45,627 stars are a different instrument that cannot be read as usage volume.

• https://pypistats.org/api/packages/pymilvus/recent recorded 2026-08-14

  last_month = 5,976,463 downloads for pymilvus (last_week 1,484,328, last_day 241,978).

• https://github.com/milvus-io/milvus recorded 2026-08-13

  Repo page, 45,627 stargazers. No download, deployment or customer figure of any kind.

• https://zilliz.com/what-is-milvus recorded 2026-08-13

  "Open-source vector database built for billion-scale vector similarity search"; LF AI & Data trademark attribution in the footer. No published adoption count.

A complete, durable, cluster-capable retrieval system, which is the 4 rung on the storage feature matrix, which bands on how much of the persist-index-retrieve path a product provides itself, at what scale and over how many retrieval modes. One band below vespa: Milvus scales horizontally with compute and storage separated and combines dense vector search with BM25 full text, but the 5 rung is for a platform that hosts and evaluates ranking or embedding models inside the serving path, and the recorded evidence for Milvus does not show that. Independent ANN-Benchmarks coverage and the first-party VDB Bench harness are what keep the basis a benchmark rather than a feature read. Rebanded on this matrix when the product moved out of agent_tools_protocols, where it had been the anchor for a three-product vector-database cluster; the band did not move.

• https://github.com/milvus-io/milvus recorded 2026-08-18

  Repo page - vector database feature set, standalone and distributed deployment.

• https://milvus.io/ recorded 2026-08-18

  Milvus 3.0 announcement headline; VDB Bench named as the benchmarking harness; index guidance from HNSW through DiskANN at billion scale; BM25 full-text search.