Releasesby GlyphPriority pay-per-token for Foundation Model APIs
Priority pay-per-token, also referred to as priority mode, for Foundation Model APIs pay-per-token is generally available for partner models and in preview for open source models. Priority mode is for latency-sensitive, real-time applications that need more consistent availability under load.
Also shipped Jul 22, 2026Databricks in July 2026 →
External OAuth and configurable scopes for MCP servers (General availability)
Two new hierarchical parameters let you bind MCP servers to your organization’s existing identity provider (IdP) and control which OAuth scopes MCP clients discover. This enables MCP clients such as Claude, Cursor, and ChatGPT to authenticate against your corporate IdP instead of Snowflake OAuth.
OpenAI API adds hard spend limits for organizations and projects
A monthly cap makes affected API requests return 429 once tracked spend reaches the limit. Spend alerts can warn before traffic is interrupted.
AI Gateway now supports streaming transcription
AI Gateway now supports streaming transcription . Previously, transcription required a complete audio file and returned the full transcript in a single response. Now you can stream audio in as it's captured and get transcript updates back as the model produces them, keeping latency low for uses like live captioning and voice input.
Extension version pinning in CREATE or ALTER EXTENSION deprecated
Specifying an explicit version in CREATE or ALTER EXTENSION is deprecated. From 2026-08-05 the version clause is ignored and the extension's default version is installed, with a warning.
Transcription on OpenRouter
OpenRouter has introduced an audio transcription API at /api/v1/audio/transcriptions, supporting base64 audio input and returning JSON text.
Volumetric Policies for Adaptive Traces are Generally Available
Cost-effectively get the right representative mix of traces with volumetric policies! Now Generally Available, volumetric policies intelligently balance cost, service representation, and value to automagically give you the right traces. We’ve bolstered volumetric policies under the hood to be even more cost predictable.
More from Agent Bricks and Mosaic AIEvery Agent Bricks and Mosaic AI release →
Google Gemini 3.6 Flash now available as a Databricks-hosted model
Model Serving now supports Google Gemini 3.6 Flash as a Databricks-hosted model. You can access this model using Foundation Model APIs . Customers are responsible for ensuring their compliance with applicable terms .
Google Gemini 3.5 Flash Lite now available as a Databricks-hosted model
Model Serving now supports Google Gemini 3.5 Flash Lite as a Databricks-hosted model. You can access this model using Foundation Model APIs . Customers are responsible for ensuring their compliance with applicable terms .
Anthropic Claude Opus 5 now available as a Databricks-hosted model
Databricks Model Serving now supports Anthropic Claude Opus 5 as a Databricks-hosted model. To access this model, use: Foundation Model APIs pay-per-token Query reasoning models Query vision models
Connect and govern any external model provider
Connect and govern any external model provider (such as OpenAI, Anthropic, Amazon Bedrock, and newly released models like Meta's Muse Spark 1.1) through Unity Gateway. Register a provider once as a governed Unity Catalog securable, and every team can query it through Unity Gateway with full governance and observability.
MLflow trace storage in Unity Catalog will soon be available by default for workspaces with the compliance security profile enabled
MLflow trace storage in Unity Catalog tables will be available by default for workspaces with the compliance security profile enabled in mid-August 2026. Store MLflow traces in Unity Catalog tables for scalable, governed trace storage: storage is unlimited, traces are stored in OpenTelemetry (OTel) format, access is governed through Unity Catalog schema and...
MLflow trace storage in Unity Catalog is now generally available
MLflow trace storage in Unity Catalog tables is now generally available. Databricks recommends storing traces in Unity Catalog for new and production workloads: storage is unlimited, traces are stored in OpenTelemetry (OTel) format, access is governed through Unity Catalog schema and table permissions, and you can query traces with SQL, Genie, AI/BI...
Google Gemini 3 Pro is retired
Google Gemini 3 Pro is retired. Gemini 3.1 Pro is the recommended replacement. See Deprecated and retired models for migration guidance.
Upcoming general availability of Unity Gateway
Unity Gateway, the Databricks governance solution for enterprise AI, is scheduled for general availability soon. Built on Unity Catalog, Unity Gateway extends governance to the runtime interactions between models, agents, MCP servers, and tools.