Source
Mar 17, 2025: Document AI release notes
Mar 17, 2025: Document AI release notes When you review the answers that the Document AI model provides, you can now highlight the answer within a document by selecting the locate answer icon.
May 08, 2025: Document AI updates
The release of a new version of the foundational Arctic - TILT model in Document AI includes improvements in checkbox identification. These improvements are available to new Document AI model builds.
Jul 07, 2025: Account Usage: New CREDENTIALS view
Jul 07, 2025: Account Usage: New CREDENTIALS view You can now view information about the following types of credentials in the new CREDENTIALS view in the ACCOUNT_USAGE schema: Programmatic access tokens Passkeys Time - based one - time…
Jun 03, 2025: Snowflake Copilot inline (Preview)
Snowflake announces the preview of Snowflake Copilot inline, an expansion of the existing Snowflake Copilot experience that gives you the ability to query Snowflake Copilot from within your SQL code.
Feb 14, 2025: Document AI release notes
The release of a new version of the foundational Arctic - TILT model in Document AI includes improvements in the following areas: Checkbox identification Answers to yes/no questions Overall model quality across internal and external…
Sep 23, 2025: AI_FILTER Performance Optimization (Preview)
AI_FILTER includes a performance optimization that delivers a 2 - 10x speedup and reduces token usage by up to 60% for suitable queries using AI functions in SELECT, WHERE, and JOIN … ON clauses.
2025_06 Bundle (Disabled by default)
Introduced in the 9.27 release (September 8 - 10, 2025) as Disabled by Default; account admins can enable for testing. Status planned to change in October 2025 to Enabled by Default; however, this schedule is subject to change.
2025 Performance improvements
The following performance improvements were introduced in 2025. Released Description Impact September 2025 More efficient workload distribution. Improves query execution time by detecting and adaptively redistributing workloads across…