Snowflake (NYSE: SNOW) announced dynamic model routing within its Cortex AI Gateway, along with plans to add two open-source AI models to its platform, according to a company press release.

The dynamic model routing feature automatically selects an AI model based on task complexity, cost, and quality requirements. Lower-complexity tasks are directed to more efficient models, while tasks requiring deeper reasoning are routed to frontier models. The feature is integrated into Snowflake's flagship AI products, including Snowflake CoCo and Snowflake CoWork, and is also available to third-party AI agents using Cortex AI Gateway.

Snowflake said internal testing showed that agents using dynamic model routing built a dbt pipeline with up to three times greater token efficiency compared to a frontier-model-only approach, while maintaining the same quality. A separate test showed engineering teams completed the same number of pull requests with 25% greater token efficiency. The company noted that results may vary based on workload and configuration.

Snowflake also said it plans to add DeepSeek-V4-Flash 0731 and GLM-5.3 to its Cortex AI platform, both currently in or approaching private preview. The company's internal testing, using the ADE-bench evaluation framework, showed DeepSeek-V4-Flash scored 74.4% on data engineering tasks, while GLM-5.2 scored 62.8% using fewer tokens than other models tested.

The Cortex AI Gateway, first announced in July 2026, also includes tools allowing administrators to track token usage, set per-user quotas, allocate costs to teams, and establish spending limits across AI applications and agents.

"Enterprises are becoming much more rigorous about the economics of AI," said Sridhar Ramaswamy, CEO of Snowflake. "Snowflake's role is to absorb that complexity so customers can focus on outcomes while we optimize model choice underneath."

Snowflake's model library includes models from Anthropic, OpenAI, Google, SpaceXAI, Meta, and Mistral, in addition to the newly announced open models.