Snowflake to add dynamic model routing to Cortex AI Gateway

Snowflake said it is adding dynamic model routing to its Cortex AI Gateway as it adds DeepSeek and Z.ai models to its ecosystem.

The new features in Cortex AI Gateway are designed to improve AI economics, expand model choices and give enterprises more granular AI usage and costs.

Model routing is becoming a critical feature to avoid runaway AI costs and token spend. In May, Boomi outlined plans to buy Lunar.dev to route models to align costs and tasks. Nvidia released NeMo Switchyard, an open model routing technology for AI agents, along with Nemotron 3.5 Lightning. Stripe is reportedly looking to buy OpenRouter. Simply put, model routing will be part of your AI architecture.

• Snowflake Summit 2026: Redrawing the Boundary Between Data, Context, and Action
• Snowflake Summit 2026: Context, custom model training, Iceberg V3
• Snowflake expands AWS partnership, acquires Natoma, delivers strong Q1

Snowflake said dynamic model routing will be entering private preview shortly. The addition of model routing rounds out Cortex AI Gateway, which was launched in July as a platform for governing agents and optimizing them.

According to Snowflake, dynamic model routing will do the following:

• Automatically select a model based on quality and cost for a task.
• Dynamic model routing will be integrated across Snowflake CoCo and Snowflake CoWork and extended to third party AI agents via Cortex AI Gateway.
• Routine tasks will be routed to cheaper models and complex work will go to frontier models.
• Snowflake said it will add DeepSeek-V4-Flash 0731 and GLM-5.3 as choices in its model library.

Snowflake said its internal testing shows that using a mix of open and proprietary models can improve efficiency by 3x.

• Constellation ShortList™ Metadata Management & Data Governance
• Constellation ShortList™ Semantic & Context Management for Analytics and AI
• Constellation ShortList™ Automated Cloud Data Warehouse Services

In a blog post, Snowflake CEO Sridhar Ramaswamy said:

"Companies need to be asking: Are we moving faster? Are we lowering costs? Are we creating better products and customer experiences? Ultimately, AI has to translate into measurable business value.

I refer to this as intelligence efficiency, or how effectively a company turns compute, models, data, and context into business impact. More revenue. Lower costs. Faster execution. Achieving this efficiency requires model flexibility, both in the freedom to use the best models for the job and in the ability to leverage intelligent routing to match each task to the right model based on cost, speed, and performance. The industry has spent the last several years optimizing the capabilities of individual models. Now, enterprises need to optimize the system around them for outcomes."