Milvus

What Milvus does

Milvus is an open-source vector database founded in 2017 and maintained by Zilliz, headquartered in San Jose, California. It is purpose-built for storing, indexing, and querying high-dimensional vector embeddings at scale, supporting dense, sparse, binary, and multimodal vectors alongside scalar fields. The platform runs in three deployment modes sharing a unified API: Milvus Lite (embedded Python library for notebooks and edge devices, installed via pip), Milvus Standalone (single-machine Docker deployment), and Milvus Distributed (Kubernetes-based cluster for billion-scale workloads).

The technology stack includes 10+ index types (HNSW, IVF, Product Quantization, GPU), hardware-accelerated compute via AVX512, Neon SIMD, and NVIDIA GPU indexing, plus the new Loon storage engine and Vortex columnar format underpinning the upcoming Milvus 3.0 and Vector Lakebase releases. Milvus 2.6.x delivers sub-10ms latency at billion-scale, 100x faster metadata filtering, and 7x faster full-text search than Elasticsearch, with three-layer tiered storage reducing storage costs by up to 87%.

Milvus monetizes through Zilliz Cloud, a fully managed commercial service offered in serverless and dedicated cluster tiers on AWS, Google Cloud, and Microsoft Azure, with BYOC options for regulated enterprises. Revenue mechanics combine freemium OSS adoption (Apache 2.0 license, 44.9K+ GitHub stars) with subscription and usage-based paid tiers, plus consulting-led enterprise deployments via system integrators such as Accenture and Deloitte. The customer base spans hundreds of enterprises across e-commerce (Walmart, eBay, Shopee, Tokopedia, Airbnb), social media (Reddit, Roblox, Line), financial services (PayPal, ZipRecruiter), healthcare (Doximity, OpenEvidence), legal (Filevine, Rexera), security (Palo Alto Networks, Trend Micro), and AI-native startups (Read AI, Notta, Exa, Credal AI).

Name

Milvus

Legal name

Milvus

Website

https://milvus.io

Company type

Private

Founded year

2017

Operating status

Operating

Headcount range

1–10 employees

Ownership category

akta.pro rank

Product category

Vector Database

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182), Computer Systems Design and Related Services (54151)

SIC

Services-Computer Programming, Data Processing, Etc. (7370)

akta.pro primary industry

Data Security & Privacy Managed Services (DLP/Encryption) (BPAEADAL)

akta.pro secondary industry

Security Operations Center (SOC) as a Service (BPAEADAB)

Keywords

• Vector database software
• Similarity search engine
• Open-source database
• AI embeddings storage
• Vector search platform

Headquarters

HQ city

Redwood City

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

1. Open-source software (free): Milvus is an open-source vector database distributed under the Apache 2.0 license. Organizations can download, deploy, and use Milvus at no cost on their own infrastructure (Milvus Lite, Standalone, or Distributed modes). This drives community adoption and ecosystem growth without direct revenue from the open-source core.
2. Zilliz Cloud (managed Milvus): Zilliz Cloud is the fully managed commercial service for Milvus, available in serverless and dedicated cluster tiers on AWS, Google Cloud, and Microsoft Azure. It provides a SaaS offering with BYOC (bring-your-own-cloud) options for security and compliance requirements. Zilliz Cloud generates subscription and usage-based revenue from organizations seeking zero-ops vector database deployments.
3. Vector Lakebase (Zilliz Cloud exclusive): Vector Lakebase is a new Zilliz Cloud-exclusive product that extends cloud-resident vector search with batch analytics, interactive discovery, and external data lake connectivity. It represents an additional premium commercial tier for organizations needing lake-native vector capabilities beyond standard Milvus deployments.

Pricing tiers

Go-to-market motion3 records

Distribution channels7 records

Marketing channels12 records

Core offering

Milvus is an open-source vector database purpose-built for storing, indexing, and searching high-dimensional vector embeddings used in GenAI applications. It scales elastically to tens of billions of vectors with sub-10ms query latency and is offered through three deployment modes (Lite, Standalone, Distributed) plus Zilliz Cloud, the fully managed commercial service available across AWS, GCP, and Azure.

Product overview

Milvus is an open-source vector database platform designed for GenAI applications, available in three deployment modes: Milvus Lite (embedded Python library for notebooks and edge devices), Milvus Standalone (single-machine Docker deployment for small-to-medium scale), and Milvus Distributed (Kubernetes-based cluster for billion-scale enterprise workloads). The platform is complemented by Zilliz Cloud, a fully managed cloud service offering serverless and dedicated cluster options with 10x performance improvement over self-hosted deployments. Recent additions include Vector Lakebase (extending vector search with lake-native storage and zero-copy data lake connectivity) and memsearch (open-source persistent memory library for AI agents). The ecosystem includes management tools like Attu (GUI), Milvus Backup, Deep Searcher, and Claude Context. All deployment modes share a unified Python/Go/Java/Node.js/C# API, with native integrations for LangChain, LlamaIndex, DSPy, Haystack, Ragas, OpenAI, and Hugging Face.

Differentiator

Problem solved

Functional benefit

Products and services

• Milvus Open-source vector database built for GenAI applications. Scales to tens of billions of vectors with sub-10ms query latency. Supports multiple deployment modes (Lite, Standalone, Distributed) sharing a unified API under Apache 2.0 license.
• Zilliz Cloud Fully managed Milvus cloud service available in serverless and dedicated cluster tiers across AWS, Google Cloud, and Microsoft Azure. Offers 10x faster performance than self-hosted Milvus with SaaS and BYOC deployment models for enterprise security and compliance.
• Vector Lakebase Zilliz Cloud-exclusive product that extends vector search with lake-native storage, batch analytics, interactive discovery, and external data lake connectivity. Features a zero-copy semantic data plane allowing the same vectors to serve production queries, discovery sessions, and training-data pipelines without data migration.
• memsearch Open-source library that gives AI agents persistent, long-term memory across conversations powered by the Milvus vector database. Stores all agent memories as human-readable plain-text files without vendor lock-in.

Quantifiable outcome

• Sub-10ms query latency at billion-scale vector search workloads
• +7 more outcomes

Named customers62 records

Segments4 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration10 records

AI capability10 records

Feature12 records

Partnerships

15 partnerships are on record, tiered core, supporting and strategic.

• LangChaincoreTechnology or IntegrationMilvus is deeply integrated with LangChain as a native vector store retriever. The langchain-milvus package provides seamless RAG pipeline construction, hybrid search, full-text search, asynchronous search, and Agent RAG capabilities. This integration is a primary go-to-market channel for reaching LangChain's large developer community building LLM applications.
• LlamaIndexcoreTechnology or IntegrationMilvus integrates with LlamaIndex through the llama-index-vector-stores-milvus package, enabling RAG, async API, full-text search, hybrid search, and metadata filtering. LlamaIndex's large user base of developers building LLM applications serves as a distribution channel for Milvus adoption.
• OpenAIcoreTechnology or IntegrationMilvus provides official integration with OpenAI's Embedding API for generating vector embeddings, documented in tutorials showing semantic search workflows using OpenAI embeddings with Milvus as the vector store. OpenAI's dominant position in the LLM market makes this integration critical for developer accessibility.
• Hugging FacecoreTechnology or IntegrationMilvus integrates with Hugging Face's transformers, datasets, and sentence embedding models for generating vector representations. Official documentation includes a question-answering tutorial using Hugging Face SQuAD dataset and sentence-transformers models with Milvus for semantic search.
• DSPycoreTechnology or IntegrationMilvus is integrated into the DSPy framework through the MilvusRM retriever module, enabling DSPy programs to leverage Milvus' vector search capabilities for RAG optimization. The dspy-ai[milvus] package provides one-line installation for this integration, targeting Stanford NLP's DSPy user community.
• Haystack (deepset)coreTechnology or IntegrationThe milvus-haystack integration enables Haystack users to use MilvusDocumentStore and MilvusEmbeddingRetriever for building RAG pipelines with Haystack's open-source Python framework. deepset, Haystack's parent company, provides official support and co-marketing for this integration.
• RagassupportingTechnology or IntegrationMilvus integrates with Ragas for evaluating RAG pipeline performance. Official documentation shows how to use Ragas metrics to assess answer quality, faithfulness, and context relevance for RAG systems built on Milvus and OpenAI.
• MemGPTsupportingTechnology or IntegrationMemGPT integration with Milvus enables long-context memory management for AI agents built with the MemGPT framework, expanding Milvus' reach in the AI agent memory use case.
• Anthropic (Claude Code)coreTechnology or IntegrationZilliz released memsearch ccplugin as a dedicated persistent memory plugin for Anthropic's Claude Code AI coding assistant. This plugin uses Milvus-powered Memsearch to give Claude Code agents persistent, human-readable memory across conversations, extending Milvus into the AI coding assistant ecosystem.
• Linux Foundation (Vortex)strategicTechnology or IntegrationVector Lakebase builds on Vortex, a Linux Foundation open-source columnar file format. This open-source standard ensures interoperability and positions Milvus/Zilliz within the broader data lake ecosystem, supporting Lance, Iceberg, Parquet, and Vortex table formats.
• Kioxia CorporationsupportingTechnology or IntegrationKioxia Corporation integrated its KIOXIA AiSAQ open-source software technology into Milvus 2.6.4, enabling SSD-optimized vector search that addresses DRAM scalability bottlenecks in high-volume inference and RAG workloads.
• NVIDIAsupportingTechnology or IntegrationMilvus supports GPU-based indexing as part of its index offerings, leveraging NVIDIA GPUs for hardware-accelerated vector operations. The multimodal RAG pipeline uses NVIDIA GPUs alongside Milvus for efficient processing.
• AccenturesupportingImplementation/ SI/ Consulting PartnerAccenture leverages Zilliz Cloud (fully managed Milvus) to power semantic retrieval across training content, knowledge assets, and coaching context for enterprise learning and workflow modernization engagements.
• DeloittesupportingImplementation/ SI/ Consulting PartnerDeloitte uses Zilliz Cloud to power semantic retrieval across knowledge, policy, and client workflow context for enterprise digital transformation and AI implementation projects.
• CoreWeavesupportingGTM or Marketing PartnerCoreWeave published technical documentation on deploying Milvus (alongside Dragonfly and Pinecone) on CoreWeave's cloud infrastructure for production-ready RAG for agentic AI systems, enabling high-performance vector search on GPU-optimized cloud infrastructure.

Scale indicators15 records

Recent moves6 records

Expansion highlights5 records

Direct peers

• Pinecone: Pinecone is a leading fully managed vector database competing directly with Zilliz Cloud. Both target enterprise AI applications requiring production-grade vector search, and Pinecone is one of the most prominent alternatives cited by Milvus users during evaluation.
• Weaviate: Weaviate is an open-source vector database with strong hybrid search capabilities, offering both self-hosted and managed cloud deployment. It competes with Milvus on the same RAG, semantic search, and GenAI use cases among developer and enterprise audiences.
• Qdrant: Qdrant is an open-source vector search engine written in Rust with a managed cloud offering. Reddit notably selected Milvus over Qdrant based on performance and scalability, indicating direct head-to-head competition for production vector search workloads.
• Chroma: Chroma is an open-source embedding database widely adopted by AI developers for prototyping and RAG applications. It competes with Milvus Lite in the developer/PLG segment and is a common comparison point cited by Milvus in its own positioning.

Broad incumbents

• Elasticsearch (Elastic): Elasticsearch has added dense_vector field types and hybrid search capabilities, positioning itself as a broad incumbent for organizations seeking vector search within an existing search/analytics platform. Milvus 2.6.x benchmarks itself 7x faster than Elasticsearch on full-text search.
• MongoDB: MongoDB Atlas Vector Search adds vector capabilities to its document database platform. As a broad incumbent with massive enterprise install base, MongoDB represents a bundling threat for organizations already standardized on its ecosystem.

Emerging players

• Vespa.ai: Vespa is a search and serving engine with native vector and hybrid search capabilities, used for large-scale retrieval applications. It overlaps with Milvus in use cases like recommendation, semantic search, and RAG, particularly among internet-scale deployments.
• pgvector: pgvector is an open-source PostgreSQL extension that adds vector similarity search to the world's most popular open-source database. It competes with Milvus for developers who prefer to stay within their existing PostgreSQL stack rather than adopt a dedicated vector database.
• LanceDB: LanceDB is a serverless vector database built on the Lance columnar format, targeting AI applications with embedded and cloud-native deployment models. It overlaps with Milvus Lite and Milvus Distributed in the developer and lake-native vector segments.
• Marqo: Marqo is a tensor and vector search engine combining ML model inference with vector storage. It competes with Milvus in multimodal and end-to-end AI search workloads, particularly for cloud-native deployments targeting developers.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/milvusio
• GitHubgithub.com/milvus-io/milvus
• YouTubeyoutube.com/@MilvusVectorDatabase/playlists
• LinkedInlinkedin.com/company/the-milvus-project
• Facebookfacebook.com/io.milvus.5

Revenue estimate

Valuation estimate

Number of profiles

Subsidiaries6 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Milvus