Zilliz announced the release of Milvus 3.0, a major architectural update to its open-source vector database, introducing lake-native data access and a more expressive retrieval engine for production AI applications. The update enables organizations to build and serve production-grade indexes over vector data in object storage and open data formats, and expands Milvus beyond nearest-neighbor search to support ranking, aggregation, sparse search, and multi-vector retrieval. The release advances the shared technical foundation behind open-source Milvus and Zilliz Cloud, with Zilliz Cloud delivering a fully managed Vector Lakebase. James Luan, Co-founder and CTO of Zilliz, said Milvus 3.0 brings production retrieval closer to where data lives and provides a more expressive engine for modern AI applications. The lake-native architecture allows indexing and retrieval to operate directly over data in object storage. External Collections define Milvus collections over data in Lance, Iceberg, Parquet, or Vortex, building indexes without copying source data, with incremental synchronization keeping indexes current. Loon, the new manifest-based storage engine, reduces read amplification for low-latency point access and uses Vortex as its default storage format. Snapshots create point-in-time, read-only views of live collections at minimal incremental storage cost. A new Spark connector exposes Milvus as a Spark DataSource V2, allowing Spark, Databricks, and EMR pipelines to read from and write to Milvus