Zilliz

Zilliz is a data infrastructure company built by the creators of the open-source Milvus vector database. Its commercial offering, positioned as a "Vector Lakebase," combines real-time vector serving, iterative discovery and batch analytics over a single source of truth, with the company stating support for scales of 100 billion-plus entities and more than 10,000 queries per second at consistent latency.

The platform provides full-spectrum search across vector, text, JSON and geospatial data, with hybrid retrieval, metadata filtering and reranking, including BM25 full-text search that can be combined with vector search. Real-time serving is organized into tiers — performance-optimized, capacity-optimized and tiered-storage — alongside massive multi-tenancy with unlimited namespaces, hot/cold data serving, and global multi-region clusters with replication and failover. An on-demand compute mode charges per query rather than per provisioned capacity, supports online backfill and schema evolution without disrupting serving, and includes an optional access mode that builds and serves indexes on Zilliz while customer data remains in their own S3 buckets in Iceberg, Lance, Vortex or Parquet formats. A command-line interface is offered for management, search and analytics, and an interactive "Zilliz Agent" guides product selection on the company site.

Business model

Zilliz sells its Milvus-based vector database as a managed cloud service, with a free entry tier, a fully managed SaaS deployment, and a bring-your-own-cloud (BYOC) option in which customer data remains inside the customer VPC while Zilliz operates the system. Milvus itself remains available as open source for self-hosted deployment, and Zilliz provides migration paths from lake data, Milvus, Elasticsearch and other vector databases.

Consumption-based pricing tied to compute and storage, including compute-unit-based serving tiers and an on-demand search mode billed per query; the site cites illustrative figures of $9.9 per 1,000 searches and $53.7 per month to store 1 billion vectors plus index.

Traction

The company reports production use across more than 10,000 enterprises over eight years and publishes customer testimonials from Exa (entity search), a company serving hundreds of thousands of clinicians, and a multilingual RAG deployment.

Latest developments

Zilliz announced the public preview of its Vector Lakebase, a fully managed, Milvus-powered service unifying real-time vector search, lake-scale discovery and AI data operations, and released an official CLI for management, search and analytics.