Zilliz Vector Lakebase

Zilliz Vector Lakebase is a Zilliz Cloud release that adds interactive discovery, batch analytics and search over external data lakes to the managed Milvus vector database. It is for AI teams with billions of vectors who today copy embeddings between a serving database and a lake, paying in time and storage.

Zilliz Vector Lakebase is a major release of Zilliz Cloud, the managed service from the company behind the open-source Milvus vector database, announced in public preview in June 2026. It keeps Zilliz Cloud's real-time vector search at the core and adds three ways to work on the same data: interactive discovery, large-scale batch analytics, and search directly on external data lakes. Underneath is a unified lake-native storage layer built on the Vortex format, and the external lake search reads Lance, Iceberg, Parquet and Vortex tables in a zero-copy mode, leaving source data where it is and syncing incrementally on refresh. Serving is split into Performance-Optimized, Capacity-Optimized and Tiered-Storage tiers, and a new On-Demand Search option bills compute pay-as-you-go; Zilliz's own benchmark puts one On-Demand workload at $318 a month against $4,937 for a comparable serverless product, a vendor figure rather than an independent test. The problem it targets is that moving billions of vectors between a serving database and an analytics system can take days, so many teams skip the feedback loop. Vector Lakebase runs alongside Zilliz Cloud's Serverless, Dedicated and BYOC options in more than 30 regions on AWS, Google Cloud and Microsoft Azure. Zilliz names Zillow, OpenEvidence, Exa, Filevine and MiniMax among more than 10,000 enterprise and AI team users, and reports Milvus at over 44,000 GitHub stars and 100 million Docker pulls. Milvus is Apache 2.0 licensed and graduated from the LF AI & Data Foundation in June 2021.

Category

Data & Analytics

Pricing

Freemium, Usage-based, Contact for pricing

Target Market

CTOs, Data Engineers, ML Engineers, Enterprise Developers

Deployment

Cloud-first, Multi-cloud

Customers

10,000+ enterprises and AI teams (vendor figure, June 2026)

• ✓
  Tiered real-time serving

  Performance-Optimized, Capacity-Optimized and Tiered-Storage tiers let teams trade latency against storage cost for each collection.

• ✓
  On-Demand Search

  Pay-as-you-go compute for search over large collections, aimed at workloads that do not justify always-on capacity.

• ✓
  External data lake search

  Zero-copy search over Lance, Iceberg, Parquet and Vortex tables, with incremental sync on refresh and source data left in place.

• ✓
  Lake-native unified storage

  A shared storage layer built on Vortex holds one logical copy of the data for serving, discovery and batch analytics.

• ✓
  Batch analytics on the same data

  Large-scale batch processing runs next to real-time serving without exporting vectors into a separate analytics system.

• ✓
  Multi-cloud deployment

  Serverless, Dedicated and BYOC options in more than 30 regions on AWS, Google Cloud and Microsoft Azure.

• •
  Re-indexing after an embedding model change

  A team re-embeds and analyses a billion-vector corpus against the lake copy instead of exporting it from the serving cluster for days.

• •
  Search over an existing lakehouse

  Embeddings already written to Iceberg or Parquet tables become searchable in place, skipping a bulk import into a vector database.

• •
  Data curation for model training

  Interactive discovery and batch analytics over the same vectors let ML teams find duplicates or clusters before building a training set.

• •
  Cost-tiered production retrieval

  Hot collections stay on Performance-Optimized serving while long-tail data sits on Tiered-Storage, lowering spend on rarely queried vectors.

Zilliz Cloud has a perpetual free cluster (5 GB, 2.5M vCUs a month) and $100 of trial credits for 30 days, extended by another $100 and one year when a card is added. Serverless and Dedicated bill on usage, On-Demand Search is pay-as-you-go compute, and BYOC is quoted by sales. Unpaid organisations are frozen when credits lapse. The free-tier and credit terms come from docs.zilliz.com, read 2026-10-07; the pricing page did not render, so per-unit rates are not stated here.

This page was written from 5 sources, 4 on domains other than zilliz.com.

1. 1.01net.it — zilliz launches vector lakebase extending the worlds most ad
2. 2.docs.zilliz.com — free trials
3. 3.hn.algolia.com — hn.algolia.com
4. 4.en.wikipedia.org — Milvus (vector database)
5. 5.zilliz.com — zilliz launches vector lakebase unified data platform for aivendor