“Enterprises are becoming much more rigorous about the economics of AI”: Snowflake wants to help you cut AI costs by choosing the right model for the right task

Snowflake has announced new dynamic model routing capabilities for flagship products along with expanded access to open-weight models for customers.

The new capabilities build on Cortex AI Gateway, according to the firm. It was unveiled in July this year and provides customers with a “unified foundation” for agent governance and routing requests.

The aim here, the company noted, is primarily focused on optimizing AI token consumption, which has become a recurring pain point for enterprises over the last 12 months.

Snowflake said the new capabilities mean Cortex AI Gateway will “automatically select" a specific model based on the quality and cost for particular tasks.

Put simply, this will see Snowflake assign more efficient models for lower-complexity or repetitive tasks. Meanwhile, tasks that require “deeper reasoning” are routed to frontier models.

This means customers can reduce unnecessary inference spend and remove the need for laborious manual model selection for each individual task, according to the firm.

Snowflake CEO Sridhar Ramaswamy said the update comes in direct response to growing concerns over AI-related costs. Token costs have surged so far across 2026, while a host of major providers have switched to new consumption-based models, resulting in increased costs for enterprises.

“Enterprises are becoming much more rigorous about the economics of AI. The question is no longer how much AI they are using, but whether that AI is translating into meaningful business value,” said Ramaswamy.

“Achieving intelligence efficiency requires the flexibility to use the best model for each task as the landscape evolves. Snowflake’s role is to absorb that complexity so customers can focus on outcomes while we optimize model choice underneath.”

Dynamic routing will also be integrated across Snowflake CoCo and Snowflake CoWork, and will be made available for third-party AI agents.

Snowflake expands model library with open-weight options

In addition to model routing capabilities, Snowflake revealed it will expand access to several leading open models, such as DeepSeek-V4-Flash 0731 and GLM-5.3.

The addition of these models adds to what is already an extensive library available for customers, Snowflake said. Enterprises can already choose from a variety of models from Anthropic, Google, Meta, Mistral, and OpenAI.

“This portfolio gives customers more freedom to choose the right combination of performance and cost for each workload,” the company said in a blog post.

Once again, the focus here is firmly on helping to manage costs by providing a broader range of options.

Internal testing conducted by the firm found that enterprises using a combination of open and proprietary models can “deliver comparable quality while materially improving efficiency”.

In one evaluation detailed by the company, agents using dynamic model routing through Cortex AI Gateway were able to build a data build tool (dbt) pipeline with up to three times greater token efficiency compared to a frontier model-only approach – all while maintaining the same quality.

“In a separate test, engineering teams completed the same number of pull requests with 25 percent greater token efficiency,” the company said in a blog post.

FOLLOW US ON SOCIAL MEDIA

Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.