What's New

AI

Claude Sonnet 5.5 as the default AI model

Sigma now uses Claude Sonnet 5.5 by default for AI features with Snowflake, Anthropic, Databricks, or Amazon Bedrock as the AI provider.

Sigma agents and formula assistant continue to use Claude Sonnet 5. If your organization does not permit or have access to Claude Sonnet 5.5, Claude Sonnet 5 is used instead.

For more information, see Supported AI models

Verify and understand Sigma Assistant answers (Beta)

To verify an answer before using it in a decision or presentation, you can now review contextual citations and analysis breakdowns in Sigma Assistant.

• Contextual citations link values in a response to the query results that support them. Hover over a cited value to preview its sources, filters, and calculation, or select it to inspect the query results.
• Analysis breakdowns explain how Assistant answered your question in plain language. Review the sources, filters, and calculations, then inspect the underlying SQL and data lineage for more detail.

For more information, see Ask natural language queries with Sigma Assistant.

AI apps

Sigma tables (Beta)

Use Sigma tables to create use, and edit warehouse-native tables directly in Sigma. Sigma tables are backed by database tables in your data platform, so edits you make in Sigma are automatically written to the underlying database table. This makes every Sigma table an editable data source, with local Sigma tables being available within the workbook containing it, and organization-wide Sigma tables being shareable across multiple Sigma documents (including workbooks, data models, and reports) and external tools connected to your data platform.

Because Sigma tables are writable, multiple users can contribute to the same table, making Sigma tables the foundation for collaborative apps that capture structured input, manage reference data, and build editable datasets shared across teams and systems.

For more information, see Sigma tables overview.

API

Manage workbooks and reports as code (GA)

You can use the Sigma API to retrieve, update, and create workbooks and reports based on a JSON or YAML representation. Using this representation with the Sigma API, you can manage and review workbooks and reports without using the Sigma UI.

The following endpoints can now be used to manage workbooks based on a code representation:

• Get workbook (GET v2/workbooks/:workbookId)
• Create workbooks (POST v2/workbooks)
• Update a workbook from a code representation (PUT v2/workbooks/:workbookId/contents)

The following endpoints can now be used to manage reports based on a code representation:

• Get report (GET v2/reports/:reportId)
• Create reports (POST v2/reports)
• Update a report from a code representation (PUT v2/reports/:reportId/contents)

For more information, see Manage workbooks as code and Manage reports as code. For examples of various features represented as code, see the Workbook representation example library and Report representation example library.

New API endpoints for custom email templates for workbook exports

The following endpoint is now available to find custom email templates for workbook exports in Sigma:

• List export email templates for an organization (GET /v2/organizations/settings/exportEmailTemplates)

To apply a custom template to email exports, set configV2.exportEmailTemplateId when calling any of these endpoints:

• Add workbook schedule (POST /v2/workbooks/{workbookId}/schedules)
• Update a workbook schedule (PATCH /v2/workbooks/{workbookId}/schedules/{scheduleId})
• Export a workbook (POST /v2/workbooks/{workbookId}/send)

New API endpoints for managing AI budget assignments (Beta)

The following endpoints are now available to manage monthly AI budget assignments for users, teams, and account types in Sigma:

• List AI budget assignments (Beta) (GET /v2/organizations/settings/aiBudgets/assignments)
• Assign an AI budget (Beta) (POST /v2/organizations/settings/aiBudgets/assignments)
• Delete an AI budget assignment (Beta) (DELETE /v2/organizations/settings/aiBudgets/assignments/{assignmentId})

You can set monthly budgets and the priority of team budget assignments programmatically. These endpoints require API credentials owned by a user assigned the Admin account type. For more information, see Set monthly AI budgets (Beta).

Revoke user sessions and API credentials

The Revoke a user’s access (POST /v2/members/{memberId}/revoke) endpoint now invalidates active Sigma sessions and deletes API and embed credentials, in addition to revoking OAuth refresh tokens.

Invalidating a session for a user logs them out of Sigma on their next request, and embed URLs signed with the deleted credentials stop working. You can’t revoke access for the user who owns the API credentials used to make the request.

Warnings for unsupported workbook and report content

The Get a workbook (GET /v2/workbooks/{workbookId}) and Get a report (GET /v2/reports/{reportId}) endpoints now include a warnings array in responses to requests with includeContents=true. The warnings identify content in the document that Sigma has dropped from the response because it is not supported in the code representation. Review these warnings before using the representation to create or update a document.

Admin

Audit logging for AWS Europe (Zurich)

Audit logging is now supported for AWS Europe (eu-central-2).

See Access and explore audit logs.

Audit logging for GCP United Kingdom (London)

Audit logging is now supported for GCP United Kingdom (europe-west2).

See Access and explore audit logs.

Audit logging for AWS Asia Pacific (Mumbai)

Audit logging is now supported for AWS Asia Pacific (ap-south-1).

See Access and explore audit logs.

Configure a BigQuery connection to use OAuth (GA)

Authenticating to BigQuery from Sigma using connection-level OAuth is now generally available.

For more information, see Connect to BigQuery with OAuth.

Monthly AI budgets (Beta)

To monitor and limit AI spend, admins can now set a default monthly budget per user and override it for specific account types, teams, or individuals. Customize model prices to reflect your negotiated rates, and review each user’s spend and budget usage in AI settings.

Users can view their own spend and budget in their profile and see warnings in chat as they approach their limit. After a user reaches their budget, Sigma blocks subsequent AI requests.

Budgets cover Sigma agents, Sigma Assistant on the home page and in workbooks, explain this chart, and the formula assistant. AI column and warehouse agent usage don’t count toward these budgets. Budgets apply per user across supported features.

For more information, see Set monthly AI budgets (Beta).

Bug fixes and improvements

• Charts generated by Snowflake Cortex Agents now display correctly in Sigma Assistant.

• Previously, when validating the code representation of a document, the Invalid kind error identified the parent element kind. Now, it identifies the offending field in the element instead.

• The AI disclaimer in themed embeds now has a transparent background so it blends into the surrounding chat area.

• The List workbooks (GET /v2/workbooks) endpoint would sometimes return a “Payment Required” error if the skipPermissionCheck parameter was set to true.

• When you create or update a workbook from its code representation, Clear control actions that target a modal, drawer, or popover no longer cause an “unknown page” validation error.

• When you use a table from a tagged data model as a data source for a Sigma agent, the agent configuration now displays the version tag.

• You can now reauthorize MCP tools used by agents from Profile > Integrations making it easier to keep using tools with agents that you use from automated actions or the API.