Databricks releases, every cloud
- Metastore-level ABAC policies are in Beta
Applies across every catalog in the metastore. Replaces per-catalog policy setup.
- Session restore for serverless jobs is in Public Preview
Restores Python variables and Spark session into a new notebook for debugging or exploration, requires eligible serverless compute job runs.
- ABAC DENY policies are in Beta
Explicitly deny MANAGE ACCESS CONTROL privilege, taking precedence over grants.
- Zerobus Ingest into default storage is now in Public Preview
Supports tables backed by default storage, requires Lakeflow Connect setup.
- Pipeline events system table is in Beta
Records Lakeflow pipeline event logs, capturing lifecycle transitions and errors, for querying and building alerts.
- Private network gateway is in Private Preview
Connects serverless compute to VPC and on-premises systems through a managed gateway, reusing existing network connectivity configuration.
- Run the change, do not just read it. Hands-on labs in your own Databricks workspace, graded when you submit. Browse labs
- The INSERT, UPDATE, and DELETE privileges are in Beta
Require Databricks Runtime 18.1 or above, alternative to MODIFY privilege.
- Alert system tables are in Public Preview
Contains alerts and alert_evaluation_history tables for auditing and monitoring.
- C++ SDK for the Zerobus Ingest connector is now available in Beta
Supports JSON and Protocol Buffers for ingesting into Unity Catalog Delta tables.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Role-based access control (RBAC) is in Public Preview
Lets users assume a role with limited permissions. Requires account and workspace admin setup.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Reddit Ads connector is now available in Beta
Ingests advertising campaign data and performance metrics from Reddit Ads into Databricks.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Custom URLs for your Databricks account (Public Preview)
Replaces default URL, requires claim and enable process.
- Thinking Machine Labs Inkling now available as a Databricks-hosted model in Public Preview
Supports text and image inputs with a 1 million token context window. Access via Foundation Model APIs.
- Restrict access to AI Functions with Unity Catalog permissions (Public Preview)
Requires account team enablement, restricts access to task-specific AI Functions.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - MATCH_RECOGNIZE is now in Beta
Requires enabling Match Recognize preview in workspace settings.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Metric view parameters are now in Public Preview
binds values at query time, replacing separate views per variant.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Block specific internet destinations in network policies (Public Preview)
Blocks destinations regardless of network access mode, enforced through REST API.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Managed agent memory is in Beta
Stores memory in Unity Catalog memory stores, accessed through REST API.
- Get this table by email. One Monday mail covering the week, filtered to the clouds and products you run. Weekly digest
- Omnigent coding agent meta-harness is now in Beta
Wraps coding agents like Claude Code and Codex with a common platform. Swap runtimes with one line of configuration.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - External data access for pipeline streaming tables and materialized views (Public Preview)
Enables external Delta and Iceberg clients to read data without copying. Uses Unity Catalog and Iceberg catalog REST APIs.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - UNIQUE constraint support is now in Public Preview
Enables join elimination and DISTINCT simplification on Photon-enabled compute. Applies to Unity Catalog Delta Lake tables.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Real-time mode in Spark Declarative Pipelines on Lakeflow, and the update_flow API are now available (Public Preview)
Achieves end-to-end latency as low as 5ms, uses update output mode.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Spark Declarative Pipelines on Lakeflow parameters are now in Beta
Define key-value pairs at pipeline level, referenced in SQL using named parameter syntax.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Configure Microsoft Entra ID SSO for Power BI (Public Preview)
Uses account-level federation policy, replacing manual config.
- SQL alert task in Lakeflow Jobs is now in Public Preview
Evaluates a Databricks SQL alert as part of a Lakeflow Job, returning its state as output.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Alibaba Cloud Qwen3.5 122B A10B now available as a Databricks-hosted model in Public Preview
Accessible via Foundation Model APIs, available in Public Preview.
- 5X-Large SQL warehouse size (Public Preview)
Provides 512 workers, available in all regions, controlled from Previews page.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Lakeflow Designer is now in Public Preview
Offers drag-and-drop canvas and natural language for building data workflows.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Query-based connectors for Lakeflow Connect (Public Preview)
Ingests data from databases using a cursor column, supporting Oracle, Teradata, and others.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - ai_classify and ai_extract now support v2 signatures (Public Preview)
Supports improved quality and performance with native composability with ai_parse_document.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Classification and Information Extraction are in Public Preview
Built on v2 signatures of ai_classify and ai_extract, they provide visual UIs for document analysis.
- Inbound Private Link for performance-intensive services (Public Preview)
Enables private connectivity for Zerobus Ingest and Lakebase Autoscaling.
- Customer-managed keys for Unity Catalog (Public Preview)
Protects data with user-owned encryption keys, replacing automatic encryption.
- HITRUST compliance controls (Public Preview)
Manages risk and demonstrates security and privacy compliance. Requires enabling in settings.
- Serverless outbound IPs available through public JSON endpoint (Public Preview)
Replaces existing stable IPs, requires Public Preview enrollment.
- JAR tasks on serverless compute is now in Public Preview
Allows running JAR jobs without a cluster.
- Qwen3-Embedding-0.6B now available in Public Preview as a Databricks-hosted model
Accessible via Foundation Model APIs on a pay-per-token basis.
- Configure default Python package repositories for Spark Declarative Pipelines on Lakeflow (Public Preview)
Allows installation from internal Python repositories without index-url or extra-index-url. Requires workspace admin configuration.
- Managed MCP servers is now Public Preview
Allows AI agents to connect to Databricks resources and external APIs.
- MySQL connector in Lakeflow Connect (Public Preview)
Enables incremental data ingestion from various MySQL databases. Requires access request through account team.
Headlines, dates and product areas are Databricks' own, from the release notes published for each cloud, and every item links to the note it came from. The one-line summaries are ours. Not a Databricks product and not affiliated with Databricks, Inc.