Databricks on AWS
1201 releases that have landed on AWS, newest first.
- Databricks Runtime 17.1 and Databricks Runtime 17.1 ML are in Beta
Includes improvements to streaming, SQL, and connectors, with reliability and performance enhancements.
- Google Gemma 3 12B now available on Mosaic AI Model Serving
It supports text inputs and both real time and batch inference workloads.
- Simplified compute form enabled by default
Applies to all-purpose and jobs compute in the Databricks UI.
- The CAN VIEW permission on SQL warehouses is now generally available
Allows viewing query history and profiles without running queries.
- Must be metastore admin to transfer share ownership for Delta Sharing
Replaces share owner transfer ability, requires metastore admin role.
- Serverless notebooks: Restore Python variables after idle termination
Restores Python variables from snapshot after idle termination. Requires reconnection to restore.
- Run the change, do not just read it. Hands-on labs in your own Databricks workspace, graded when you submit. Browse labs
- Databricks connector for Power BI now supports the ADBC driver (Public Preview)
Replaces ODBC driver, requires configuration change.
- Enhanced Security and Compliance add-on now available to all customers with usage-based pricing
Includes compliance security profile and enhanced security monitoring, with charges based on feature usage.
- Git support for alerts
Tracks alert changes, requires placing alerts in a Databricks Git folder. Newly cloned alerts are paused.
- Dashboard visualization updates
Fixes area chart error and swaps latitude and longitude in Point Maps.
- Genie interface, query, and data improvements
Editable filters display multiple values, Genie suggests popular SQL queries from notebooks.
- Databricks Apps support for Lakebase resources
Adds Lakebase database instances as app resources, accessible as environment variables.
- Parent tasks (Run job and For each) now have a separate limit
Separate limit doesn't count against overall limit, see Resource limits.
- Dashboard theme settings and page enhancements
Retains theme settings on import/export, increases max pages to 15.
- Databricks SQL version 2025.20 is now available in Preview
Includes new features and behavioral changes in Preview channel.
- Genie visualization and permissions improvements
Respects SQL sort order by default. Allows CAN MANAGE users to delete Genie Agents.
- Git folders now supports multiple Git credentials per user (Public Preview)
Managed through the UI, supports multiple Git providers.
- Bring your own data lineage (Public Preview)
Includes external assets and workflows run outside of Databricks. Requires Unity Catalog.
- Get this table by email. One Monday mail covering the week, filtered to the clouds and products you run. Weekly digest
- Dashboard Git integration and workflow improvements
Diff reviews show queries and text across multiple lines, and Genie supports multistatement SQL queries.
- Genie API enhancements and SQL function improvements
Adds list, delete conversations, and delete Genie Agents endpoints. Improves SQL function application to relevant questions.
- New Databricks JDBC Driver is open source
Licensed under Apache 2.0, accepting community contributions.
- File arrival triggers with file events is now generally available
Requires file events enabled on external locations like S3 or Azure storage. Raises trigger limit to 1,000 per workspace.
- MLfLow support for compliance security profile standards
Supports HIPAA, PCI-DSS, and FedRAMP Moderate standards.
- Parse documents using ai_parse_document (Beta)
Extracts structured content from unstructured documents using Databricks Foundation Model APIs.
- Use PDFs in Information Extraction and Custom LLM (Beta)
Converts PDFs to Markdown, storing results in a table for input in Information Extraction and Custom LLM.
- Databricks Runtime 17.0 is GA
Replaces previous runtimes, with end of support dates listed.
- OIDC federation for Databricks-to-open Delta Sharing is generally available
Uses JSON web tokens from recipients' IdP for authentication.
- Custom rate limits for model serving endpoints
Limits apply to overall traffic, users, service principals, or user groups.
- Migrate Spark Declarative Pipelines on Lakeflow pipelines from legacy publishing mode
Replaces single catalog and schema publishing with multi catalog and schema support. Defaults to multi catalog mode after migration.
- Sharing managed Iceberg tables in Delta Sharing is in Public Preview
Supports Databricks-to-Databricks and open sharing. Requires adding tables to a share.
- Expanded feature support for FIPS-Compliant environments
Includes FedRAMP Moderate and IRAP Canada Protected B support. Available in serverless and standard compute.
- File upload and theme customization features
Uploads files to Unity Catalog from dashboards. Customizes themes with font and color settings.
- Genie filter editing and audit log enhancements
Editable filters refine responses for string columns with entity matching. Audit log adds comment_type field for feedback comments.
- AI documentation for Unity Catalog tables and columns now uses the same model as Databricks Assistant
Uses a Databricks-hosted model, staying available even if Partner-powered AI features are turned off.
- Jobs & Pipelines in the left navigation menu
Replaces Pipelines and Workflows items, unifying their functionality.
- Moving streaming tables and materialized views between pipelines is in Public Preview
Supports moving tables created by Spark Declarative Pipelines on Lakeflow.
- Attribute-based access control (ABAC) in Unity Catalog is in Beta
Uses tags and user-defined functions to enforce fine-grained access controls.
- Automatic liquid clustering is now GA
Enables intelligent clustering key selection on Unity Catalog managed tables.
- Dashboard visualization fixes and schedule updates
Fixes tooltip highlighting and table collapse issues. Parameterized schedules are temporarily disabled.
- AI Gateway is now generally available
Streamlines AI model usage and management with features like permission and rate limiting, and payload logging.