Snowflake Direct Share Now Supports ML Model Sharing Across Accounts

Snowflake has quietly extended its Direct Share feature to support machine learning models, enabling ML models registered in the Snowflake Model Registry to be shared across accounts. This update meaningfully expands the options available to teams using Snowflake as an ML platform. Sharing behavior differs depending on whether inference is run in a warehouse or on Snowflake Container Services (SPCS), requiring separate model versions for each target platform. On the provider side, ML models can be shared either by granting privileges directly to a share object or through a database role, with the latter approach recommended for easier role structure replication on the consumer side. The feature was noted as having shipped without broad announcement, highlighting the importance of monitoring Snowflake's documentation for incremental updates.