Zilliz announced the public preview of Zilliz Vector Lakebase, an extension of its Milvus‑based vector database that adds lake‑native storage and on‑demand compute. The new platform keeps real‑time vector search at its core while enabling interactive discovery, large‑scale batch analytics, and direct search on external data lakes—all from a single logical data copy. By unifying these workloads, Vector Lakebase addresses the AI‑system loop that modern teams run: serve results, collect feedback, mine and refine data, then serve again. The preview is now available on Zilliz Cloud across more than 30 AWS, Google Cloud, and Azure regions, giving enterprises a globally distributed foundation without the need for separate serving, exploration, and batch‑processing stacks.
Zilliz Launches Vector Lakebase on Zilliz Cloud
The release pairs Zilliz Cloud’s production vector search engine—used by companies such as Zillow, OpenEvidence, Exa, Filevine, MiniMax, and over 10,000 enterprises—with a shared, lake‑native data foundation. Vector Lakebase introduces three additional workload modes: interactive discovery, batch analytics, and external data lake search. All workloads operate on the same logical dataset, and compute is billed only when active, eliminating the cost of idle servers.
Charles Xie, Founder and CEO of Zilliz, said, “Production vector search is and will remain at the heart of what Zilliz does … Vector Lakebase is what we believe comes next: 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.”
Robert Guo, VP of Product at Zilliz and a Milvus architect, added that teams “asked for a way to keep their data in one place and run very different workloads against it — from real‑time agent memory to overnight semantic deduplication.” Vector Lakebase delivers this through a unified storage layer on the Vortex format, tiered serving for production, and on‑demand compute for other tasks. The platform’s design reflects the industry shift from isolated query‑only services to continuous AI loops, where serving, learning, and data preparation must happen on the same data plane to avoid costly data movement and latency.
Core Capabilities and Performance Claims
Vector Lakebase offers five integrated capabilities:
- Tiered Real‑Time Serving – Three production tiers (Performance‑Optimized, Capacity‑Optimized, Tiered‑Storage) deliver 1,000+ QPS with single‑digit‑millisecond latency down to 10–50 QPS with ~100 ms latency, all backed by a 99.99 % uptime SLA and cross‑region high availability.
- On‑Demand Search – Pay‑as‑you‑go compute billed only for active usage. In Zilliz’s internal benchmark on one billion 768‑dimensional vectors with 10 hours of monthly compute, on‑demand search cost $318 versus $4,937 for a comparable serverless path (≈ 1/15 the cost).
- External Data Lake Search – Zero‑copy “External Collection” mode adds indexing and full‑spectrum search to existing Lance, Iceberg, Parquet, and Vortex tables, with incremental sync on refresh.
- Full‑Spectrum AI Search – Supports dense and sparse vectors, text, JSON, and geospatial data, with hybrid retrieval, BM25, regex, multi‑vector, iterative search, and reranking via Cohere, Voyage AI, RRF, and weighted/boost/decay strategies.
- Unified Lake‑Native Storage – Built on Vortex, an open columnar format optimized for random reads, paired with object‑storage‑aware indexes that cut read amplification by over 90 %. A 100‑million‑row schema backfill typically completes in single‑digit minutes without disrupting active queries.
Together, these features aim to replace parallel serving clusters and separate batch systems with a single platform that maintains consistent indexes, versioned data, and compute that scales to zero between jobs.
Availability and Early‑Access Incentives
Vector Lakebase entered public preview on Zilliz Cloud on the same day as the announcement. It is offered alongside Serverless, Dedicated, and BYOC deployment options, spanning more than 30 regions on the three major cloud providers. New work‑email sign‑ups receive $100 in free credits at zilliz.com. Organizations that currently run serving, discovery, and analytics on separate stacks can request a tailored walkthrough from Zilliz.
Key Takeaways
- Zilliz launched the public preview of Vector Lakebase, extending Milvus‑based vector search with lake‑native storage and on‑demand compute.
- The platform introduces tiered real‑time serving, on‑demand search (costing $318 in Zilliz’s benchmark versus $4,937 for a comparable serverless path), external data lake search, full‑spectrum AI search, and unified Vortex storage.
- Vector Lakebase is available now in public preview on Zilliz Cloud across 30+ AWS, Google Cloud, and Azure regions, with $100 free credits for new work‑email sign‑ups.
TechInsyte's Take
Vector Lakebase consolidates serving, discovery, and analytics workloads onto a single data foundation, which could simplify architecture for AI teams that currently stitch together multiple stacks. The cost claim for on‑demand search is notable, though it reflects an internal benchmark; real‑world pricing will depend on workload patterns and data volume. Buyers should evaluate the maturity of the Vortex format and the zero‑copy external collection feature against existing lake solutions before committing to a production rollout.
Source: Businesswire
