Zilliz Vector Lakebase

Zilliz Vector Lakebase is a unified AI data platform that combines real-time vector search, interactive discovery and large-scale batch analytics on a single lake-native data foundation, aimed at data and AI platform teams running retrieval at enterprise scale. It extends Milvus — the open-source vector database with 44,000+ GitHub stars — into a managed Zilliz Cloud service with a 99.99% uptime SLA.

Zilliz announced Vector Lakebase on 10 June 2026 as a major Zilliz Cloud release that pairs its production vector database with a shared, lake-native data foundation, eliminating the copies and parallel stacks that typically separate serving from analytics. The architecture keeps vector data primarily in cloud object storage and builds on Vortex, an open Arrow-compatible columnar format used as the default storage layer for Loon, the manifest-based storage engine in Milvus 3.0 — a design Zilliz says cuts read amplification by more than 90% for low-latency access on object storage. Five capabilities define the platform: Tiered Real-Time Serving across Performance-Optimized, Capacity-Optimized and Tiered-Storage production tiers with a 99.99% uptime SLA; On-Demand Search with pay-as-you-go compute that Zilliz's internal benchmarks put at roughly one-fifteenth the cost of serverless alternatives; External Data Lake Search offering zero-copy indexing over existing Lance, Iceberg, Parquet and Vortex tables; Full-Spectrum AI Search spanning vectors, text, JSON and geospatial data in hybrid retrieval; and Unified Lake-Native Storage so on-demand and batch jobs bill only when compute is active. Founder and CEO Charles Xie describes it as "one data foundation where the same vectors can serve a production query, anchor a discovery session, and power a multi-petabyte training-data pipeline — without copies, migration, or a parallel stack." Vector Lakebase is in public preview on Zilliz Cloud across more than 30 regions on AWS, Google Cloud and Microsoft Azure, alongside Serverless, Dedicated and BYOC deployment options, with $100 in credits for new work-email signups. Zilliz, headquartered in Redwood City, California, is the creator of Milvus — 44,000+ GitHub stars, 100M+ Docker pulls, Apache 2.0 licensed and a graduated LF AI & Data project — and says it serves more than 10,000 enterprises and AI teams including Zillow, Salesforce, OpenEvidence, Exa, Filevine, MiniMax and Read AI.

Category

Data & Analytics

Pricing

Usage-based, Contact for pricing

Target Market

CTOs, CIOs, Data Engineering Leaders, Data Scientists, Enterprise Developers

Headquarters

Redwood City, United States

Customers

10,000+ enterprises and AI teams

• ✓
  Tiered Real-Time Serving

  Three production tiers — Performance-Optimized, Capacity-Optimized and Tiered-Storage — backed by a 99.99% uptime SLA.

• ✓
  On-Demand Search

  Pay-as-you-go compute that Zilliz internal benchmarks place at roughly 1/15 the cost of serverless alternatives.

• ✓
  External Data Lake Search

  Zero-copy indexing and search directly on existing Lance, Iceberg, Parquet and Vortex tables without duplicating source data.

• ✓
  Full-Spectrum AI Search

  Hybrid retrieval across vector, full-text, JSON and geospatial data in a single query path.

• ✓
  Unified lake-native storage on Vortex

  An open Arrow-compatible columnar format underpinning Milvus 3.0's Loon engine, cutting read amplification by over 90%.

• ✓
  Flexible deployment

  Available as Serverless, Dedicated and BYOC across 30+ regions on AWS, Google Cloud and Microsoft Azure, with self-hosted Milvus on Kubernetes, Docker or air-gapped environments.

• •
  Production RAG and semantic search

  Serves low-latency vector queries for retrieval-augmented generation against a single logical copy of enterprise data.

• •
  Multi-petabyte training-data pipelines

  Runs large-scale batch analytics over the same vectors that serve production traffic, with no migration between systems.

• •
  Exploratory data discovery

  Interactive discovery sessions query the same lake-native store used for serving, billed only when compute is active.

New signups with a work email receive $100 in credits. On-demand and batch jobs bill only when compute is active; specific tier rates are not published. Milvus itself remains free and Apache 2.0 licensed for self-hosting.

This page was written from 2 sources, 2 on domains other than zilliz.com.

1. 1.prnewswire.com — zilliz launches vector lakebase extending the worlds most ad
2. 2.blocksandfiles.com — 5273617