Databricks releases, every cloud
- Python UDTFs in Unity Catalog are now generally available
Runs on serverless compute, classic compute, and SQL warehouses. Requires Databricks Runtime 18.
- Anthropic Claude Opus 5 now available as a Databricks-hosted model
Access via Foundation Model APIs, pay-per-token.
- Automated setup for S3 external locations
Uses AWS IAM temporary delegation to configure external locations and provision required resources.
- Email notifications for expiring service principal tokens are now generally available
Sends emails to workspace admins 7 days before token expiration for used tokens with lifetime over 7 days. No configuration is required.
- Priority pay-per-token for Foundation Model APIs
Priority mode is for latency-sensitive applications, requiring partner or open source models.
- Context-based ingress control for workspaces public access is now generally available
Uses allow and deny rules to control access based on identities, network sources, and request types.
- Run the change, do not just read it. Hands-on labs in your own Databricks workspace, graded when you submit. Browse labs
- ai_extract and ai_classify are now available by default for workspaces with the compliance security profile enabled
Replaces manual text extraction and classification workflows, requires compliance security profile.
- Data Classification is now available by default for additional compliance standards
Requires compliance security profile with C5, TISAX, or IRAP controls.
- Databricks SQL alerts are now available by default for workspaces with the compliance security profile enabled
Supports all compliance security profile standards, monitors data and KPIs via scheduled queries.
- Anthropic Claude Sonnet 4 is scheduled for retirement
Replaced by Claude Sonnet 4.6, retiring October 9, 2026.
- Zhipu AI GLM 5.2 now available as a Databricks-hosted model
Supports 1 million token context window and long-horizon reasoning. Accessed via Foundation Model APIs.
AWS, read it on their docsAzure(not on this cloud)GCP, read it on their docsSAP, read it on their docs - Budgets are now generally available
Define spending thresholds with optional per-user overrides and actions like email alerts.
- Anthropic Claude Sonnet 5 now available as a Databricks-hosted model
Accessible via Foundation Model APIs, requiring compliance with applicable terms.
- Data Classification is now available by default for some workspaces with the compliance security profile enabled
Requires compliance security profile and specific controls like HIPAA or HITRUST.
- ai_query is now generally available
Lets you query AI models from SQL or Python.
- Faster inference tables for custom model serving endpoints
Ingests through endpoint telemetry, with configurable sampling, and requires AI Gateway setup.
- ai_classify is now generally available
Classifies text content according to custom labels with support for multi-label classification.
- ai_extract is now generally available
Extracts structured data from text and documents according to a provided schema. Supports nested objects and arrays.
- Get this table by email. One Monday mail covering the week, filtered to the clouds and products you run. Weekly digest
- Endpoint telemetry for custom model serving endpoints is generally available
Persists OpenTelemetry logs, traces, and metrics to Unity Catalog tables.
- Anthropic Claude Fable 5 now available as a Databricks-hosted model
Requires pay-per-token Foundation Model APIs.
- Identity federation enabled by default for all new workspaces
Does not affect existing workspaces.
- External lineage is now generally available
Registers external assets like Salesforce or MySQL sources and Tableau dashboards. Automatically records source lineage for Lakeflow Connect pipelines.
- Inbound Private Link for performance-intensive services is now generally available
Supports Lakebase Autoscaling and Zerobus Ingest.
- Foundation model Unity Catalog permissions is generally available
Requires account team enablement, controls access to system.ai schema.
- Column popularity is now available in the Catalog Explorer
Shows column query count in the Overview tab, determined by read queries.
- Anthropic Claude Opus 4.8 now available as a Databricks-hosted model
Access via Foundation Model APIs, pay-per-token.
- Automatic identity management with Microsoft Entra ID is now GA
Replaces SCIM provisioning, syncs users and groups from Microsoft Entra ID.
AWS, read it on their docsAzure(not on this cloud)GCP, read it on their docsSAP, read it on their docs - Unity Catalog managed Apache Iceberg tables, foreign Apache Iceberg tables, and Apache Iceberg v3 features are generally available
Supports liquid clustering, predictive optimization, and materialized views. Requires Databricks Runtime or external Iceberg-compatible engines.
- ai_parse_document is now available by default for workspaces with the compliance security profile enabled
Requires compliance security profile and specific controls like HIPAA and PCI-DSS. Parses structured content from unstructured documents.
- Catalog commits are now generally available
Enables multi-statement transactions on Unity Catalog managed Delta tables, supported by products like MLflow and Delta Sharing.
- Automatic identity management
Syncs users and groups from identity providers like Microsoft Entra ID and Okta.
AWS, read it on their docsAzure(not on this cloud)GCP, read it on their docsSAP, read it on their docs - Azure UK South now has a dedicated regional control plane
New workspaces require firewall allowlist entries for SCC relay tunnel and other IPs. Existing workspaces use UK West control plane unless migrated.
- Governed tags are generally available
Allow administrators to define controlled tags for Unity Catalog and workspace objects, enforcing consistent metadata tagging.
- ABAC policy functions now use snake_case naming
Replaces camelCase with snake_case, old forms still work but will be deprecated.
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.