What is DuckLake?

DuckLake is an open table format for data lakehouses, created by DuckDB Labs, built around a simple but distinctive idea: storing a lakehouse's table metadata (schema, snapshots, file listings, statistics) in a proper SQL database rather than in the large number of small JSON and Avro metadata files that formats like Apache Iceberg and Delta Lake typically rely on.

CREATE DATABASE my_lakehouse (TYPE ducklake);

Reaching version 1.0 (a stable, backward-compatible specification) in mid-2026, DuckLake provides the features expected of a modern lakehouse format, schema evolution, time travel, ACID transactions, and works with the actual data files still stored as Parquet in object storage, only the metadata layer is fundamentally different. It's MIT licensed, designed to be vendor-neutral, and supports being read and written by multiple query engines, not exclusively DuckDB.

The core argument DuckDB Labs makes for this design is that SQL databases are already very good at managing many small, concurrent metadata operations, exactly what a lakehouse catalog needs, and that reusing that maturity avoids some of the operational complexity that comes with coordinating consistency across thousands of small metadata files scattered across object storage.