Databricks Release Notes

The product known as DLT is now Spark Declarative Pipelines on Lakeflow. No migration is required to use Spark Declarative Pipelines on Lakeflow. See Spark Declarative Pipelines.

Managed Apache Iceberg tables are now in Public Preview. You can read from and write to these tables from Databricks or from external Iceberg engines using the Iceberg REST Catalog API. These tables are integrated with predictive optimization to apply advanced optimizations, including liquid clustering automatically. See What is Apache Iceberg in Databricks? and Unity Catalog managed tables for Delta Lake and Apache Iceberg.

Foreign Apache Iceberg tables are now in Public Preview. You can read Iceberg tables managed by foreign catalogs, such as HMS, Glue, and Snowflake Horizon Catalog, using Lakehouse Federation. These tables support Unity Catalog advanced features such as fine-grained access controls, lineage, and auditing. See What is Apache Iceberg in Databricks? and Work with foreign tables.

ALTER TABLE ... SET MANAGED is now available in Public Preview for participating customers. This command enables seamless conversion of Unity Catalog external tables to managed tables. It allows you to take full advantage of Unity Catalog managed table features, such as enhanced governance, reliability, and performance. See Convert external or foreign Delta Lake tables to Unity Catalog managed tables.

MLflow 3.0 is now generally available.

MLflow 3.0 on Databricks delivers state-of-the-art experiment tracking, observability, and performance evaluation for machine learning models, AI applications, and agents on the Databricks Lakehouse. See Get started with MLflow 3 for models.

Deployment jobs are now available in Public Preview.

Deployment jobs allow you to automate tasks like evaluation, approval, and deployment whenever a new model version is created, integrating seamlessly with Unity Catalog models and Lakeflow Jobs. See MLflow 3 deployment jobs.

Selecting the serverless performance setting for jobs and pipelines is now generally available.

When the Performance optimized setting is enabled, your workload is optimized for faster startup and execution time. When disabled, the serverless workload runs on standard performance mode, which is optimized for cost and has a slightly higher launch latency.

For more information, see Select a performance mode and Select a performance mode.

Databricks now supports MCP, an open standard that lets AI agents securely access tools, resources, prompt, and other contextual information using a consistent interface.

• Managed MCP servers: Use Databricks-hosted servers for easy, no-maintenance access to Unity Catalog data and tools.
• Custom MCP servers: Host your own MCP server or third-party server as a Databricks app.

See MCPs.

Cross-platform view sharing via Delta Sharing is now generally available. The data access and billing method when sharing views are updated. See How do I incur and check OpenSharing costs?.

A new system table allows you to track the shared materialized data history. See OpenSharing materialization history system table reference.

Account admins can now configure the time-to-live (TTL) of data materialization. See Configure TTL of data materialization.

You can configure and enforce outbound network policies for serverless compute resources, including SQL warehouses and model serving endpoints.

With network policies you can:

• Configure outbound access for serverless workloads.
• Allowlist specific domains and storage accounts for restricted mode.
• Enable dry-run mode to monitor policy impact before enforcement.
• View and analyze denial logs in Unity Catalog for auditing and troubleshooting.

See Manage network policies for serverless egress control

Tag policies enables admins to enforce consistent tagging across data assets such as catalogs, schemas, and tables. Tag policies define allowed tag keys and values, and control which users and groups can assign them. This enables standardized metadata management for data classification, cost tracking, access control, and automation use cases.

Tag policies are managed at the account level and apply across all workspaces. For more information, see Governed tags

You can now enable AWS PrivateLink to your S3 storage buckets from serverless compute. See Configure private connectivity to AWS-managed resources.

You can now enable AWS PrivateLink to resources in your virtual private cloud (VPC) via a network load balancer (NLB) from serverless compute. See Configure private connectivity to resources in your VPC.

Serverless GPU compute is now part of the Serverless compute offering. Serverless GPU compute is specialized for custom single and multi-node deep learning workloads. You can use serverless GPU compute to train and fine-tune custom models using your favorite frameworks and get state-of-the-art efficiency, performance, and quality.

See AI Runtime.

Databricks Apps now supports all compliance security profile standards supported by Databricks, such as HIPAA, PCI-DSS, and FedRAMP Moderate. For a full list of supported compliance standards, see Compliance security profile.

Databricks is now available in the AWS Asia Pacific (Jakarta) region (ap-southeast-3). Customers can now create new workspaces and use Databricks products in this region. This expansion allows organizations operating in or near Indonesia to deploy Databricks closer to their data and users, improving performance and meeting regional compliance requirements.

For the full list of supported regions, see Databricks clouds and regions.

• New consumer entitlement is generally available: Workspace admins can now grant consumer access as an entitlement to users, service principals, and groups. Consumer access enables limited workspace UI access, querying SQL warehouses using BI tools, and viewing dashboards with embedded or viewer credentials. See Manage entitlements and Share a dashboard.

• Dashboard schedule fix: Resolved an issue where schedules for dashboards published with global filters were not executing.
• Cross-filtering fix for histograms: Fixed an issue where cross-filtering with histograms did not work on large datasets.