Databricks releases, every cloud
- Genie Code for AI Runtime (Public Preview)
Generates distributed training code and resolves environment issues. Requires AI Runtime.
- Custom trace views in the MLflow trace explorer are in Beta
Genie generates views based on plain language descriptions, surfacing relevant trace fields and metrics.
- Web terminal on serverless GPU compute (AI Runtime) is in Public Preview
Runs shell commands, monitors GPU usage, and manages files on serverless GPU compute. Requires environment version 5 or above.
- MLflow trace storage in Unity Catalog is now generally available
Stores traces in OpenTelemetry format with unlimited storage and SQL query access.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Protobuf tensor input for custom model serving endpoints (Public Preview)
Replaces JSON with serialized KServe v2 ModelInferRequest. Requires endpoints deployed after July 9, 2026.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Custom Docker images for AI Runtime CLI workloads (Beta)
Allows custom container images for specific library versions or dependencies, replacing default images.
- Run the change, do not just read it. Hands-on labs in your own Databricks workspace, graded when you submit. Browse labs
- Improved workspace files read performance
Applies to serverless and classic compute with Databricks Runtime 18 and above, speeding up workloads like find and grep.
- AI Runtime CLI (Beta)
Submits and manages distributed training workloads on serverless GPU compute from a local machine.
- Reorganized Spark Declarative Pipelines on Lakeflow documentation
Conceptual topics are now under Concepts.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Improved workspace file write performance on classic compute
Uses asynchronous flushing by default for Databricks Runtime 17.1 and above, batching file writes and deletes.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Declarative Automation Bundles will soon default to use the direct deployment engine Coming soon
Replaces Terraform deployment engine, migrating bundles automatically on July 24, 2026.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - Databricks CLI is now GA
Stores U2M tokens in OS-native secure storage.
- Serve custom LLMs with Custom Model Serving (Beta)
Supports multimodal models and PEFT recipes, unlike Foundation Model APIs.
- Customer-managed keys now support MLflow managed evaluation features
Supports MLflow 3 scorers and requires non-CMK encrypted catalog.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - MLflow trace storage in Unity Catalog is now in Public Preview
Stores OpenTelemetry traces, governed by Unity Catalog permissions, queryable from Databricks SQL or MLflow Python SDK.
AWS, read it on their docsAzure, read it on their docsGCP, read it on their docsSAP(not on this cloud) - AI Runtime 1xH100 accelerator (Beta)
Supports 1xH100 accelerator, see Hardware options for details.
- Databricks Documentation site table of contents tabs
Replaces static sidebar with tabbed navigation, includes 6 context tabs.
- AI Runtime is now in Public Preview
Adds GPU support to serverless compute for deep learning workloads.
- Get this table by email. One Monday mail covering the week, filtered to the clouds and products you run. Weekly digest
- Declarative Feature Engineering APIs (Beta)
Define time-windowed aggregation features from data sources, materialize to Delta tables or Lakebase. Available in us-east-1 and us-west-2 regions.
- Endpoint telemetry for custom model serving endpoints (Beta)
Stores logs, traces, and metrics in Unity Catalog Delta tables using OpenTelemetry.
- Store and query MLflow traces in Unity Catalog (Beta)
Stores MLflow traces in Delta tables for long-term retention and analysis using OpenTelemetry format.
Headlines, dates and product areas are Databricks' own, and every item links to the note it came from. The one-line summaries are ours.