Snowflake AI: Dynamic Model Routing Lowers Costs

Snowflake has announced dynamic model routing within its Cortex AI Gateway and flagship AI products. This innovation aims to help businesses reduce unnecessary AI spending and improve "intelligence efficiency." The new capabilities build on the Cortex AI Gateway, which Snowflake introduced in July 2026. This gateway helps manage AI consumption and route requests. Using the same AI model for every task can increase costs, and managing many models creates more work for development teams. Snowflake's dynamic model routing addresses this by automating model selection. The system can now automatically pick the best model for a task, balancing quality and cost. This feature is integrated into products like Snowflake CoCo and Snowflake CoWork, and is available to third-party AI agents. It sends simpler tasks to more efficient models and complex tasks to more advanced ones, reducing inference spending. Snowflake is also expanding access to open models like DeepSeek-V4-Flash 07311 and GLM-5.33 through Snowflake Cortex AI. This gives customers more control over AI costs as their usage grows, helping them match each task with the right model.