augusto kiniama rosa

Posted on Jan 5 • Originally published at blog.archetypeconsulting.com on Dec 1, 2025

The Unofficial Snowflake Monthly Release Notes: November 2025

#newfeatures #data #datasuperhero #ai

Monthly Snowflake Unofficial Release Notes #New features #Previews #Clients #Behavior Changes

Welcome to the Unofficial Release Notes for Snowflake for November 2025! You’ll find all the latest features, drivers, and more in one convenient place.

As an unofficial source, I am excited to share my insights and thoughts. Let’s dive in! You can also find all of Snowflake’s releases here. This month, we provide coverage up to release 9.37 (General Availability — GA).

I would appreciate your suggestions on how to continue combining these monthly release notes. Feel free to comment below or chat with me on LinkedIn.

Behavior change bundle 2025_05 is generally enabled for all customers, 2025_06 is enabled by default but can be opted out until next BCR deployment, and 2025_07 is disabled by default but may be opted in.

What’s New in Snowflake

AI & ML Updates (Cortex, ML, DocumentAI)

Import models from Hugging Face to Snowflake (Preview), support for importing external models in preview. Besides curated Snowflake models or your own, you can bring any Hugging Face transformer to your Snowflake model registry, and use it like other models
AI_COMPLETE function (GA), Generates responses (completions) from prompts using your choice of large language model (LLM). AI_COMPLETE is the most general Cortex AI Function; it is not specialized for particular use cases like summarization or classification. It can generate various responses based on content and instructions in the prompt. Responses can be plain text or semi-structured data, with prompts containing text and images processed according to plain English instructions. Examples include explaining concepts to a child, assessing and simplifying reading levels, critiquing writing styles, estimating product ratings, comparing ads, identifying highest inflation rates in a graph, describing kitchen appliances in images, or extracting stock symbols and prices as JSON
Document Processing Playground (Preview), helps users explore Snowflake’s AI document processing via a playground interface. You can upload documents, experiment with AI_EXTRACT and AI_PARSE_DOCUMENT functions, ask questions to extract info, preview layout and OCR results, and copy SQL queries and Python code for worksheets and notebooks
Cortex Analyst Routing Mode (Preview), a query strategy that prioritizes semantic SQL and defaults to standard SQL only when needed. It simplifies SQL with guardrails from semantic views. Routing mode uses these views for higher accuracy and consistency, ensuring metrics, joins, and filters adhere to governed definitions. Benefits include: consistent metrics, safer defaults, and LLM-friendly shorter SQL
AI_REDACT function (Preview), detects and redacts personally identifiable information (PII) from unstructured text data using a large language model (LLM). AI_REDACT automatically recognizes various categories of PII (such as name, address, and subcategories like first or last name) and replaces them with placeholder values
Cortex Agents (GA), agents orchestrate data analysis from structured and unstructured sources to generate insights, plan tasks, use tools, and produce responses. They utilize Cortex Analyst and Cortex Search, along with LLMs, to analyze data through four steps: Planning, Tool use, Reflection, and Monitoring. Planning involves parsing requests, exploring options, splitting tasks, and ensuring compliance. Tools are used for data retrieval. Reflection evaluates results for next steps, while monitoring tracks performance and improves behavior
Cortex AI Functions (GA), AI Functions enable unifying structured and unstructured analytics within a single platform and accelerate intelligent application development. You can build scalable, multimodal AI pipelines inside Snowflake, supporting text, image, audio, and video intelligence without external services or data movement. These functions will soon be generally available: AI_CLASSIFY categorizes inputs into user-defined categories; AI_TRANSCRIBE transcribes audio/video files, extracting text, timestamps, and speaker info; AI_EMBED generates embedding vectors for similarity search, clustering, and classification; AI_SIMILARITY computes similarity between two inputs without creating explicit embeddings
Cortex AI_TRANSCRIBE function (GA), brings production-ready, SQL-native transcription for both audio and video content within Snowflake, making it easier to extract and analyze spoken information at scale
The general availability release includes several improvements over the preview release:
Automatic language detection improvements for higher accuracy across multilingual and mixed-language recordings.
Support for MP4 and other video files, enabling transcription and analysis of media content for advertising and sponsorship analytics.
Support for Norwegian and Hebrew, expanding language coverage to 31 languages.
Overall transcription quality improvements across diverse environments and acoustic conditions
Snowflake-managed MCP server (GA), A standards-based interface allows AI agents to securely retrieve data from Snowflake accounts without additional infrastructure. It offers standardized integration for tool discovery, comprehensive OAuth authentication via Snowflake’s built-in service, and RBAC for governance over tool discovery and invocation
Snowflake Machine Learning Experiments (Preview), allows you to gather training run info for models and evaluate via Snowsight. Create experiment runs from training parameters, metrics, and artifacts. Experiments compare data to help select the right model. Snowflake Experiments are flexible, accepting any relevant data for evaluation
Snowflake Intelligence (GA), is a powerful tool that provides insights and actionability from organizational data. Create charts, get instant answers with natural language, discover trends, and analyze data without needing technical skills or waiting for dashboards. Access and analyze thousands of data sources—structured and unstructured—simultaneously, connecting spreadsheets, documents, images, and databases

Data Lake Updates

External query engine support for Apache Iceberg™ tables with Snowflake Horizon Catalog (Preview), query Snowflake-managed Apache Iceberg™ tables with external engines supporting the open Iceberg REST protocol, like Apache Spark™. Apache Polaris™ (incubating) integrated into Horizon Catalog ensures interoperability. You can query tables in a Snowflake account using one Horizon Catalog endpoint and your existing Snowflake users, roles, policies, and authentication
Apache Iceberg™ tables: Support for bi-directional data access with Microsoft Fabric (Preview), query Snowflake-managed Iceberg tables within Fabric by connecting a Snowflake database. You can choose an existing database or set up a new one. Once connected, Fabric generates an item for accessing your Snowflake-managed tables and allows querying OneLake tables with Iceberg metadata from Snowflake. To query Fabric Iceberg tables registered in Snowflake, set up a REST catalog integration for OneLake table APIs, which supplies table information from Fabric
Replicate Snowflake-managed Apache Iceberg™ tables (Preview), this process involves copying from a source account to one or multiple target accounts within the same organization. It is seamlessly integrated with Snowflake replication and failover groups to ensure point-in-time consistency of objects on the target account

Snowflake Applications (Container Services, Notebooks and Applications, Snowconvert)

Tri-Secret Secure data protection for Snowpark Container Services block volumes (GA)
SnowConvert AI interface improvements, revised for better efficiency, control, and usability, the new interface allows running specific flows like extraction, deployment, and data validation independently. It features a dedicated project page showing available flows and provides granular control, simplifying management of complex workflows
Snowflake Native Apps support for FedRAMP on AWS for apps with containers (GA), support for FedRAMP on Amazon Web Services enables apps with containers to be distributed to any Snowflake customer within FedRAMP regions. Apps operating in FedRAMP can leverage Snowpark Container Services, including compute pools, services, jobs, and external access integrations
Improved stage volume implementation in Snowpark Container Services (GA)
Snowpark Container Services in Google Cloud (GA)

Realtime Data (Hybrid, Interactive & SnowPostGres)

Interactive tables and interactive warehouses (Preview), provide enhanced query performance and real-time data processing capabilities for your Snowflake workloads

Data Pipelines/Data Loading/Unloading Updates

Snowpipe Streaming with high-performance architecture on Azure and Google GCP (GA), enables large-scale, real-time data ingestion with high throughput and low latency into Snowflake across major cloud platforms
Snowflake Openflow — Snowflake Deployments (GA), you can now run on Snowpark Container Services (SPCS)

Data Transformations

dbt Projects on Snowflake (GA), Create, edit, test, run, and manage dbt Core projects using Snowsight Workspaces to handle project files, deploy as schema-level objects, and use SQL and Snowflake CLI for deployment in CI/CD workflows. GA features include: faster command execution—upload times reduced from 6-6.5 minutes to 40-45 seconds; no secondary roles needed for Workspaces; expanded command support—dbt docs generate and retry commands, plus more flags; enhanced Snowsight view of project DAGs and source code; improved execution and scheduling UI; easy access to artifacts for debugging and integration; and support for replicating dbt objects to failover accounts.

Security, Privacy & Governance Updates

Trust Center notifications in Snowsight (GA), enable Trust Center notifications about accounts that haven’t enrolled in multi-factor authentication (MFA)
Anomaly detection for Data Quality Monitoring (Preview), set up anomaly detection for data quality monitoring so that Snowflake automatically detects unexpected changes in the following dimensions: volume of data in a table, and frequency with which a table is being updated
Access control enhancements for cost anomalies, a cost anomaly occurs when daily consumption deviates from the expected range. Previously, only system administrators could view anomalies. Now, application roles allow you to grant some users access to view anomalies, while others can view and manage them.
Excluding objects from sensitive data classification (GA), configure Snowflake to exclude schemas, tables, or columns from automatic classification so that they are skipped during the classification process
Trust Center extensions (Preview), security partners and customers can use the Snowflake Native App Framework to create Trust Center extensions with additional scanner packages. Users can discover, install, and manage these extensions. Currently, several security partners like ALTR, Hunters, OneTrust, and TrustLogix offer Trust Center extensions in this preview
Storage lifecycle policies (GA), Schema-level objects enable managing data retention in Snowflake tables via archiving or expiring rows based on conditions. Storage lifecycle policies offer key benefits: lower storage costs by moving old data to cheaper tiers; ensure regulatory compliance through automated archival and deletion; facilitate automated data management with Snowflake-managed compute; and allow flexible data retrieval by creating tables with selected rows

Snowsight Updates

Shared Workspaces (Preview), collaborative development in Snowsight allows multiple users to work on the same files and folders within a role-based environment, enhancing team collaboration while ensuring security. Key features include creating shared workspaces, sharing files, managing access through roles, and collaborating with visibility into updates
Performance Explorer (GA), Monitor interactive metrics for SQL workloads. These metrics display the overall health of your Snowflake environment, query activity, and changes to warehouses and tables

SQL, Extensibility & Performance Updates

Enhanced Query Acceleration Service (QAS) to intelligently determine when queries with LIMIT clauses can be accelerated, more queries with LIMIT clauses, even without ORDER BY, can now be accelerated. QAS automatically identifies when this improves performance, expanding the range of queries that benefit
Preparation for renaming Snapshots feature to Backups The billing line item for the WORM snapshots feature, which is currently in preview, changes from Snapshot to Backup, affects METERING_HISTORY Account Usage view, the value of the SERVICE_TYPE column changes from SNAPSHOT to BACKUP
New DECFLOAT data type, support for the DECFLOAT data type stores numbers exactly with up to 38 significant digits and uses a dynamic base-10 exponent for large or small values. Unlike FLOAT, which approximates values, DECFLOAT represents exact values at the specified precision
You can use the DECFLOAT data type when you need exact decimal results and a wide, variable scale in the same column
Support for OAuth when authenticating with GitHub (GA), authenticate using OAuth when you’re integrating a repository on GitHub with Snowflake
Run Apache Spark™ workloads on Snowflake (GA), connect your existing Spark workloads directly to Snowflake and run them on the Snowflake compute engine. As a result, you can run your PySpark dataframe code with all the benefits of the Snowflake engine
Support for connecting Scala applications to Snowpark Connect for Spark (Preview), Connect your Scala apps to Snowpark Connect for Spark. After configuring the connection and starting the server, run Scala code to connect`
Function, CREATE OR ALTER FUNCTION, updated to support changing function definition. For example, RUNTIME_VERSION, ARTIFACT_REPOSITORY (Python), PACKAGES, IMPORTS, return type, and function body
Procedure, CREATE OR ALTER PROCEDURE, Updated to support changing procedure definition. For example, RUNTIME_VERSION, IMPORTS, PACKAGES, return type, procedure body, and ARTIFACT_REPOSITORY for Python stored procedures

Collaboration, Data Clean Rooms Updates, Marketplace, Listings & Data Sharing

Clean Rooms API Version: 11.9: fixed audience overlap threshold logic: Corrected SQL threshold comparison for accuracy. Overlap now measured as less than or equal to, with updates to private preview features
Clean Rooms API Version: 11.8: updates to private preview features
Clean Rooms API Version: 11.2: New custom templates: The following template flows have been removed from the clean rooms UI and made into custom templates that you can download and run in code: Last-touch attribution, Audience lookalike modeling, Inventory forecasting; and other updates to private preview features
Support for paid listings in the Kingdom of Saudi Arabia (KSA) (GA), providers can create paid listings in KSA. See 'Who can provide paid listings' for countries where providers can offer paid listings, and consumers can access them in KSA. For a list of countries where consumers can access paid listings, see 'Supported consumer locations'
Sharing semantic views, providers can share semantic views in private listings, in public listings on the Snowflake Marketplace, and organizational listings

Open-Source Updates

terraform-snowflake-provider 2.11.0 (This release introduces new resources for Notebooks and Semantic Views, alongside scheduling enhancements for Tasks, and fixes an Identifier Parsing bug)
terraform-snowflake-provider 2.10.1 (patch release focuses on stabilizing the authentication policy changes introduced in v2.10.0, fixed an issue with parsing the DESCRIBE output for authentication policies)
Snowflake VS Code Extension 1.20.1 (Bug Fixes — When Snowflake: Enable Native App Panel is disabled, disable any recursive directory search for snowflake.yml to improve performance in large codebases)

Client, Drivers, Libraries and Connectors Updates

New features:

Go Snowflake Driver 1.18.0 (Added validation of CRL NextUpdate for freshly downloaded CRLs, added logging of query text and parameters)
Go Snowflake Driver 1.17.1 (Added telemetry for login requests to supported platforms (such as EC2, Lambda, Azure function, and so on). You can disable the telemetry by setting the SNOWFLAKE_DISABLE_PLATFORM_DETECTION environment variable (SNOWFLAKE_DISABLE_PLATFORM_DETECTION=true), exposed QueryStatus from SnowflakeResult and SnowflakeRows in the GetStatus() function, added the CrlDownloadMaxSize parameter to limit the size of CRL downloads, added official support for RHEL9 (Red Hat Enterprise Linux 9), improved log messages, deprecated several configuration options and functions. For more information, see the Upcoming Gosnowflake v2 changes)
ODBC 3.13.0 (Added support for Decfloat types, Support cross-signed chains during OCSP check, implemented a new CRL (Certificate Revocation List) checking mechanism, enabling CRLs improves security by checking for revoked certificates during the TLS handshake process)
Snowflake Connector for Python 4.1.0 (Added official support for RHEL9, added the oauth_socket_uri connection parameter to allow users to specify separate server and redirect URIs for local OAuth server, added the no_proxy parameter for proxy configuration without using environmental variables, added the SNOWFLAKE_AUTH_FORCE_SERVER environment variable to force the driver to receive SAML tokens even without opening a browser when using the externalbrowser authentication method. The variable allows headless environments, such as Docker or Airflow) that run locally to authenticate the connection using a browser URL)
.NET Driver 5.1.0 (Added the APPLICATION_PATH to the CLIENT_ENVIRONMENT sent during authentication to identify the application connecting to Snowflake, AWS WIF (Workload Identity Federation) now also checks the application configuration and AWS profile credentials store when determining the current AWS region, added ability for users to configure the maximum number of connections by setting the SERVICE_POINT_CONNECTION_LIMIT property, added the CRLDOWNLOADMAXSIZE connection parameter to limit the maximum size of CRL (certificate revocation list) files downloaded during certificate revocation checks)
Snowflake CLI 3.13.0 (New features and updates Added the — decimal-precision global option to allow setting arbitrary precision for Python’s Decimal type, added support for the auto_suspend_secs parameter in SPCS service commands (deploy, set, unset) to configure automatic service suspension after a period of inactivity, added the snow dbt describe and snow dbt drop commands, added the snow dbt execute … retry subcommand, added the following snow dbt deploy command options: — default-target to set a default target, — unset-default-target to clear the default target, — external-access-integration to set external access integrations (needed to pull external dependencies for altering a dbt project object), — install-local-deps to install dependencies located in the project, added support for running Streamlit apps on SPCS runtime, Added grant privileges definitions to the Streamlit snowflake.yml file, updated snowflake-connector-python to version 3.18.0, Relaxed dbt profiles.yml validation rules; added extra validation for role specified in profiles.yml)
Snowflake Python API 1.9.0 (Behavior changes: Event sharing is now mandatory for all event types)
Snowpark Library for Scala and Java 1.17.0 (New features — new APIs: DataFrame.isEmpty, functions.try_to_timestamp, functions.try_to_date, functions.concat_ws_ignore_nulls, functions.array_flatten, Row.mkString (with overloads for customizable separators and formatting options), StructType.fieldNames (alias for StructType.names); Improvements : functions.when and Column.when, along with Column.otherwise, now accept any literal arguments (for example, String, int, boolean, or null) in addition to Column instances, add functions.substring overload with support for start position and length arguments, add functions.lpad overloads to pad with String, or Array[Byte], add functions.rpad overloads to pad with String, or Array[Byte], add DataFrame.sort overload with support for variadic arguments, add DataFrame.show overloads with parameters to control truncation and number of displayed rows)
Snowpark Connect for Spark 1.2.0 (Snowpark Connect for Spark New features: Relax version requirements for grpcio and aiobotocore; Improvements: Specify dependencies version in meta.yaml, build compiled and architecture-specific conda package, ensure all CloudPickleSerializer.loads are not done in TCM, include OSS SQL tests that start with the WITH clause, do not upload Spark jars when running the server for pyt, update internal queries count; Snowpark Submit Improvements: Generate unique workload names)
Snowflake ML 1.19.0 (Experiment Tracking API (snowflake.ml.ExperimentTracking module, online feature serving in Feature Store)
Snowpipe Streaming SDK 1.1.0 (With the release of the SDK version 1.1.0, Snowpipe Streaming’s high-performance architecture is now generally available for all accounts on Microsoft Azure, expanding its availability from Amazon Web Services (AWS), update on November 10, 2025: Support for Google Cloud Platform (GCP) is also added and is now generally available for all accounts)
Snowflake Connector for Google Analytics Aggregate Data 2.2.1 (Behavior changes: Event sharing is now mandatory for all event types)

Bug fixes:

Go Snowflake Driver 1.18.0 (Fixed a data race error in tests caused by the platform detection init() function, made secrets detector initialization thread safe and more maintainable)
Go Snowflake Driver 1.17.1 (Fixed a bug where GCP PUT/GET operations would fail when the connection context was canceled, fixed unsafe reflection of nil pointers on DECFLOAT function in the bind uploader, added temporary download files cleanup, added a small clarification in the oauth.go example on token escaping, ensured proper permissions for CRL cache directory,bypassed proxy settings for WIF metadata requests, fixed nil pointer dereferences while calling long-running queries, moved the keyring-based secure storage manager into a separate file to avoid the need to initialize keyring on Linux)
.NET Driver 5.1.0 (Renew idle sessions in the pool if keep alive is enabled)
ODBC 3.13.0 (Removed the trailing null termination character from the JWT header and payload)
Snowflake Connector for Python 4.1.0 (Fixed a compilation error when building from sources with libc++, added OAUTH_AUTHORIZATION_CODE and OAUTH_CLIENT_CREDENTIALS to the list of authenticators that don’t require users to set the user parameter)
Snowpark Connect for Spark 1.2.0 (Snowpark Connect for Spark: Fix tests for tcm, fix CSV column name discrepancy from Spark, use type cache for empty frames, resolve Windows OSS runner general issues; Snowpark Submit: Fix staged file reading)
Snowflake ML 1.19.0 (Model registry bug fixes: get_version_by_alias now requires an exact match of the version’s Snowflake identifier)

Conclusion

November 2025 signifies a key milestone in Snowflake’s transition from a data warehouse to a comprehensive AI Data Cloud. The main highlight is the advancement of Cortex AI, which now includes Cortex Agents and support for importing Hugging Face models. This shift expands Snowflake's capabilities from basic text generation to sophisticated, agent-driven workflows and customized machine learning processes.

However, the operational updates are just as important. The general availability of dbt Projects and the expansion of Snowpipe Streaming to Azure and GCP demonstrate Snowflake's focus on enhancing developer experience and ensuring cross-cloud compatibility. Alongside this, the interoperability of Iceberg tables with Microsoft Fabric sends a clear message: Snowflake is creating a platform that emphasizes open standards, unified governance, and smooth integration, no matter where your data or workflows are located.

Enjoy the reading.

I am Augusto Rosa, a Snowflake Data Superhero and Snowflake SME. I am also the Head of Data, Cloud, & Security Architecture at Archetype Consulting. You can follow me on LinkedIn.

Subscribe to my Medium blog https://blog.augustorosa.com for the most interesting Data Engineering and Snowflake news.

DEV Community