Maximize AI Impact: More Model Choice, Smarter Routing
Snowflake introduces dynamic model routing in Cortex AI Gateway to optimize AI workloads by selecting the most cost-effective and performant models for each task, expanding model choices with support for GLM-5.3 and DeepSeek-V4-Flash 0731.
Companies are prioritizing measurable business value from AI investments, shifting focus from usage metrics to outcomes such as revenue growth, cost reduction, and faster execution. Snowflake emphasizes intelligence efficiency, which depends on model flexibility and intelligent routing to match tasks with the right models based on cost, speed, and performance. The rise of open models has altered AI’s cost structure, making simpler tasks more economically viable for automation. This broader model choice fosters competition, innovation, and lower costs across the ecosystem.
Snowflake’s Cortex AI Gateway now offers dynamic model routing, allowing customers to define approved models and tradeoffs while the system selects the optimal model for each task. The system evaluates tasks against policies and real-time cost and performance data, then routes requests accordingly. A continuous feedback loop evaluates results and refines future routing decisions, adapting to model improvements and changing landscapes. Initial benchmarks indicate this approach delivers better economics at a given quality level compared to relying on a single model.
The new capabilities are integrated into Snowflake CoCo and CoWork, enabling users to leverage different models, agents, and workflows within a single conversational interface without managing models directly. Snowflake is expanding model choices with support for GLM-5.3 and DeepSeek-V4-Flash 0731, alongside existing models from Anthropic, Google, Mistral AI, OpenAI, and SpaceXAI. This flexibility aligns with Snowflake’s goal of helping enterprises achieve intelligence efficiency by absorbing the complexity of model selection and optimization.
Snowflake positions these updates as part of its long-standing approach to simplifying major technology shifts for customers. By integrating dynamic routing and expanded model choices into its AI platforms, Snowflake aims to help enterprises realize measurable business value from AI investments while reducing operational complexity. The company highlights that this system adapts to evolving models and workloads, ensuring continuous optimization of cost, quality, and latency.