snowflake Snowflake Blog ·

Snowflake Enhances Cortex AI Gateway with Dynamic Model Routing and Expanded Model Choice

blogaisnowflakegaarchitect
feature announcement

Snowflake has introduced dynamic model routing within its Cortex AI Gateway, enabling enterprises to automatically select the optimal AI model for each task based on cost, speed, and performance. This enhancement aims to improve "intelligence efficiency" by continuously evaluating tasks against defined policies and real-world data, optimizing AI adoption for business value. The platform also expands its model offerings with support for GLM-5.3 and DeepSeek-V4-Flash 0731, allowing greater flexibility in model choice. These capabilities are seamlessly integrated into Snowflake CoCo and CoWork, helping builders and business users leverage diverse models within a conversational interface.

  • Introduce dynamic model routing in Cortex AI Gateway
  • Expand model choice with GLM-5.3 and DeepSeek-V4-Flash 0731
  • Integrate AI optimizations into Snowflake CoCo and CoWork
Features (1)
  • Introduce dynamic model routing in Cortex AI Gateway

    Cortex AI Gateway now offers dynamic model routing, allowing customers to define model approval policies and tradeoffs. The system evaluates tasks against these policies and real-world data, selecting the best model and continuously learning from results to improve future decisions, ultimately delivering better economics at a given level of quality.

Enhancements (2)
  • Expand model choice with GLM-5.3 and DeepSeek-V4-Flash 0731

    Snowflake is expanding its portfolio of supported models, adding GLM-5.3 and DeepSeek-V4-Flash 0731. This builds on the existing choice of leading open and proprietary models from vendors like Anthropic, Google, Mistral AI, OpenAI, and SpaceXAI, providing greater flexibility and competition.

  • Integrate AI optimizations into Snowflake CoCo and CoWork

    The new dynamic model routing capabilities and expanded model choices are seamlessly integrated into Snowflake CoCo and CoWork. This allows builders and business users to leverage optimized price and performance within their existing agents and workflows through a conversational interface.

Read the original announcement →

https://www.snowflake.com/content/snowflake-site/global/en/blog/ai-intelligence-efficiency-dynamic-model-routing

Related releases