September 21, 2026: Cortex AI Function Optimization for more efficient AI implementations (Public Preview)¶

Snowflake Cortex AI Function Optimization is now available in public preview, enabling customers to automatically search across prompts and models for improved implementations of custom AI Functions. Using the AI_FUNCTION_OPTIMIZATION experiment type, customers can evaluate multiple model candidates against a consistent dataset and metric, and compare their quality and estimated cost.

Key use cases include:

• Improve AI function quality automatically: Generate and evaluate alternative prompt and model configurations without manually rewriting each implementation.
• Compare models across capability and cost tiers: Evaluate smaller, lower-cost models alongside larger, more capable models using the same data and scoring criteria.
• Identify the most cost-efficient model: Determine whether a lower-cost model meets the workload’s quality requirements or whether a more capable model provides a material improvement.
• Create an optimized AI function: Materialize the selected candidate as a reusable, governed AI function with a stable SQL interface.

After optimization, you can use the selected prompt and model directly in an AI_COMPLETE call, or create a reusable custom AI function (CREATE AI FUNCTION) from the result. The original function is not modified.

For more information, see Optimize an AI function.