Snowflake adds Gemini 4 Argon to Cortex AI in public preview

Snowflake made Google’s Gemini 4 Argon available in public preview for its customers through Cortex AI Functions on October 13. The availability gives Snowflake users a managed way to call the model through SQL and Python.

The preview widens access beyond Google’s initial restricted rollout. In its September 30 announcement, Google said access would begin with a small group of trusted cyber defenders in its Fairwind Program while it gathered feedback ahead of broader availability.

Snowflake said Argon is available across Cortex AI Functions, including through AI_COMPLETE, which generates text or image-based completions using a selected model. Snowflake attributed sustained reasoning, the ability to use visual context and support for complex, multistep enterprise workflows to Argon; the announcement did not include an independent evaluation of those claims.

Snowflake also said model calls remain within Snowflake’s security and governance perimeter. Its AI and ML documentation says models accessed through Snowflake AI Features are deployed within the Snowflake Service perimeter. Snowflake’s documentation describes preview models as intended for evaluation and subject to more frequent changes than generally available models.

Snowflake said integrations with Snowflake CoWork, Snowflake CoCo and Cortex Agents are coming soon across supported clouds through cross-region inference. It did not provide launch dates, a general-availability date, Snowflake-specific pricing or usage limits, or a list of supported cloud regions in the announcement.