All release notes

• Data metric function (DMF) FILTER clause (General availability)

The following new features and enhancements are now available in Snowflake Data Clean Rooms:

• ML Jobs in Data Clean Rooms — Generally Available: ML Jobs enables collaborators to run containerized machine learning workloads on a dedicated compute pool inside a Data Clean Room. Supported use cases include lookalike modeling, incrementality measurement, hyperparameter optimization, and distributed model training using the Snowflake distributed trainers for XGBoost, LightGBM, and PyTorch. ML Jobs are defined in a code spec and submitted through the standard template approval workflow.
• Unregister templates and data offerings: You can now remove a registered template or data offering from a registry by calling UNREGISTER_TEMPLATE or UNREGISTER_DATA_OFFERING, as long as the resource isn't currently linked to a collaboration.
• General performance improvements and bug fixes.

Artifacts in Snowflake CoWork are now generally available. An artifact is a persistent, live reference to a chart or table from an agent response that you can revisit, refresh under your own credentials, and share with a teammate who sees the same artifact filtered through their own data permissions.

...

Read more

With this release, Adaptive Compute is generally available on Amazon Web Services (AWS). You can create Adaptive Warehouses or convert existing standard warehouses to an Adaptive Warehouse without downtime.

...

Read more

The ACCOUNT_USAGE schema now includes a new CORTEX_AI_GUARDRAILS_USAGE_HISTORY view that provides visibility into the usage history of Cortex AI Guardrails.

...

Read more

Support for Azure Data Lake Storage Gen2 (Azure Data Lake Storage) for Apache Iceberg™ tables is now generally available.

...

Read more

With this preview release, Cortex Agent evaluations include two new system metrics that measure how your agent uses tools:

• Tool selection accuracy (TSA) – Whether the agent's orchestration layer invokes the tools you expect for a given query.
• Tool execution accuracy (TEA) – Whether each tool that runs receives appropriate input and returns output that meets your requirements.

...

Read more

The following updates are now available in Snowflake Data Clean Rooms:

• General performance improvements and bug fixes.
• Updates to private preview features.

• New ICEBERG_MERGE_ON_READ_BEHAVIOR parameter for Apache Iceberg™ tables (General availability)
• Deprecation: ENABLE_ICEBERG_MERGE_ON_READ

You can now set a default metadata write format at the account, database, or schema level so that CREATE TABLE and ALTER TABLE create Apache Iceberg™ tables without the ICEBERG keyword. This release includes the following capabilities:

• DEFAULT_METADATA_WRITE_FORMAT parameter to default new tables to the Snowflake table format or the Apache Iceberg™ table format.
• Automatic DDL promotion in catalog-linked databases so CREATE TABLE creates Apache Iceberg™ tables without the ICEBERG keyword.

...

Read more

Support for connecting Snowflake to Google Cloud BigLake Metastore by using an Apache Iceberg™ REST catalog integration is now generally available. You can configure workload identity federation so Snowflake authenticates to Google Cloud without long-lived service account keys, then create a catalog-linked database to discover and query BigLake Iceberg tables from Snowflake.

...

Read more

When you use catalog-vended credentials for Apache Iceberg™ tables, you can now configure outbound private connectivity so Snowflake accesses your cloud storage through a private endpoint instead of the public internet. You can enable private storage access by setting USE_PRIVATELINK_ENDPOINT = TRUE in the DEFAULT_STORAGE_CONFIG parameter of your catalog integration.

...

Read more

With this release, Snowflake storage for Apache Iceberg™ tables is now generally available. This option lets Snowflake store and manage the Iceberg table files for you, so you don't need to set up access to external cloud storage.

...

Read more

Optimized refresh is a new refresh mode for failover groups, now available in public preview. It makes account replication refreshes more efficient and predictable, and is especially valuable for customers who replicate a very large number of objects. You opt in per failover group by setting OPTIMIZED_REFRESH = TRUE on the primary failover group. The mode works with the same failover group SQL surface you already use, and replication, failover, and failback semantics are unchanged.

...

Read more

With this release, Streamlit in Snowflake in Workspaces is now generally available.

...

Read more

With this release, restricted caller's rights in Streamlit in Snowflake is now generally available. This feature requires Streamlit version 1.53.1 or later.

...

Read more

Effective June 1, Snowflake is reducing pricing for Cortex AI_TRANSCRIBE by up to 60%, making it more cost-effective to transcribe and analyze audio and video content directly in Snowflake.

...

Read more

The dynamic tables feature formerly known as the immutability constraint is now called the frozen region of a dynamic table. As part of this rename, the IMMUTABLE WHERE clause is also renamed to FROZEN WHERE.

...

Read more

Snowflake Cortex AI_COMPLETE now supports Gemini 3.5 Flash for multimodal video and audio analysis, enabling organizations to extract brand, sentiment, behavioral, and content insights from rich media directly in Snowflake. With frontier multimodal intelligence and Flash-level speed, teams can analyze video and audio content for marketing, content intelligence, compliance, and media analytics workflows without moving data outside of Snowflake.

...

Read more

You can now generate a Sensitive Data Access report to view which users accessed sensitive data in your account during a specified lookback period. The report creates a view that lists user-table-role combinations identified in query and access history for classified tables.

...

Read more

With this preview release, you can use the Governance & security » Data protection policies area in Snowsight to create data protection policies from the Dashboard, list and manage policies on the Policies tab, and view protected objects on the Objects with policies tab.

...

Read more

The following new features and enhancements are now available in Snowflake Data Clean Rooms:

• DCR Agent (public preview): You can now use a conversational Snowflake CoWork interface to join collaborations and run approved analyses including audience overlap and activation via standard_audience_overlap_v0 and standard_audience_overlap_activation_v0 templates. Learn more.
• Additional Cortex Code Skills: Users can now use Cortex Code with data clean room skills that support dropping/leaving a collaboration, linking/unlinking data offerings, and requesting new templates to be added with approval flows.

Dynamic tables now support a new refresh mode: REFRESH_MODE = ADAPTIVE.

...

Read more

When standard refresh modes don't support your transformation pattern, CUSTOM_INCREMENTAL refresh mode lets you write MERGE or INSERT logic that Snowflake executes on each refresh. You define exactly how changes reach the dynamic table while Snowflake handles scheduling, retries, and transactional guarantees. This mode supports patterns such as stream-static joins, soft-deletes, stateful aggregation, and direct migration from streams and tasks.

...

Read more

With this release, write support for Snowflake-managed Apache Iceberg™ tables by using an external query engine through Snowflake Horizon Catalog is now generally available. You can read and write to Snowflake-managed Iceberg v2 and v3 tables from any external query engine that supports the open Iceberg REST protocol, such as Apache Spark™, by using a single Horizon Catalog endpoint and your existing Snowflake users, roles, policies, and authentication.

...

Read more

• Virtual columns (General availability)

Extraction scores for the Snowflake Cortex AI_EXTRACT function are now generally available. Key capabilities include:

• Confidence scores for extracted values: Pass scores => TRUE in an AI_EXTRACT call to get back a scoring object alongside the standard response object, with a score between 0 and 1 indicating the model's certainty about each extracted value.
• Support for entities, lists, and tables: Per-field scores are returned for entity extraction, and aggregate scores are returned for list and table extraction.
• Support for fine-tuned arctic-extract models: Extraction scores work with fine-tuned arctic-extract models in addition to the default model.

...

Read more

The following updates are now available in Snowflake Data Clean Rooms:

• General performance improvements and bug fixes.
• Updates to private preview features.

The dynamic tables documentation now includes 30 new and updated pages covering the full lifecycle of building, optimizing, monitoring, and troubleshooting dynamic table pipelines.

...

Read more

Attaching a storage lifecycle policy to a dynamic table is now generally available. You can attach a storage lifecycle policy directly to a dynamic table to delete or archive rows that match a predicate, on a schedule that runs independently of dynamic table refresh.

...

Read more

Cortex AI Function Studio is now available in public preview, enabling organizations to easily create, evaluate, and optimize production-ready Cortex AI Functions for unstructured and multimodal data workflows. AI Function Studio automates prompt engineering, model selection, benchmarking, and optimization, helping teams accelerate AI application development while improving quality and reducing experimentation overhead.

...

Read more

With this preview release, you can build, validate, and register your own custom container images for use with Snowflake Notebooks and ML Jobs on Container Runtime. Custom runtime images let you include specific packages, meet compliance and security requirements, integrate with existing CI/CD pipelines, and ensure reproducibility across environments.

...

Read more

The following improvements are now available for dbt Projects on Snowflake:

• Enhanced dbt DAG with column-level lineage: The dbt DAG visualization is now significantly more capable across all three surfaces (Workspaces, dbt project object details, and Query History). Each model node displays its columns, powered by the Snowflake Horizon Catalog as the single source of truth. Selecting a column highlights all upstream and downstream models in the DAG that use that column. Additional improvements include an in-DAG search bar to find models by name, expandable upstream and downstream depth controls, a unified side panel showing row count, column count, compiled SQL, and runtime information, and an increased model display limit (up from 100 to 300). For more information, see dbt projects on snowflake manage.

...

Read more