/
New feature Beta

Managed warehouse for MLflow

Managed warehouse for MLflow is a Databricks ai / ml capability, introduced November 2026.

Lets the MLflow UI show traces stored in Unity Catalog without you picking a SQL warehouse: MLflow runs the trace queries on a warehouse it creates and manages, or on serverless compute.

  • The warehouse it creates is literally named MLflow Managed Warehouse, and it is MLflow's alone: users need no permissions on it, and cannot use it for anything else.
  • Admins can edit or delete that warehouse, and Databricks recommends, in so many words, that they don't.

Limitations: Beta, turned on from the Previews page. Works only in the MLflow UI - the MLflow Python SDK still needs a SQL warehouse. Traces must be stored in Unity Catalog, the workspace must be in a serverless region, and the dedicated warehouse needs the Lakehouse//RT Beta; without it, queries run on serverless compute.

Open in REbricked →
Category
AI / ML
Introduced
November 2026
Also known as
MLflow Managed Warehouse, Managed warehouse for MLflow traces
Verified
2026-11-05

Sources

Related in AI / ML