AI_COMPLETE

Sep 29, 2026: Response caching for Cortex AI Functions

Snowflake Cortex AI Functions now automatically caches and reuses the results of AI_COMPLETE calls. When a query makes an identical AI_COMPLETE call more than once, Snowflake can return the stored result instead of running inference again, which lowers cost and speeds up query execution.

Caching is best effort and applied automatically within a single query. Cache hits are not billed, so queries with identical rows consume fewer tokens and credits. Customers running queries with large numbers of identical rows can benefit from significant cost savings and latency reductions.

Response caching is available for AI_COMPLETE text calls. No action is required to benefit from it.

Read from the provider's text; the quote is the provider's exact lines.

Mar 05, 2026: AI_COMPLETE document intelligence (Preview)

The multimodal AI_COMPLETE AI Function now supports document inputs, providing reasoning over document files, such as PDFs or Microsoft Word files, stored in Snowflake internal or external stages. This enhancement allows AI_COMPLETE to analyze text, charts, tables, and structured data within documents. This feature joins existing support in AI_COMPLETE

Read from the provider's text; the quote is the provider's exact lines.