Snowflake Cortex

Snowflake Cortex provides comprehensive AI/ML capabilities directly within the Snowflake Data Cloud, enabling LLM operations without data movement. Cortex AI Functions (GA November 2025) include AI_COMPLETE (text generation), AI_CLASSIFY (classification), AI_FILTER (data filtering), AI_EXTRACT (entity extraction), AI_EMBED (embeddings), AI_TRANSCRIBE (audio to text), and more. Supports models from OpenAI, Anthropic (Claude), Meta (Llama 4), Google, Mistral, and DeepSeek, model selection based on task requirements. Pricing is token-based with credits varying by model (GPT-4 more expensive than Llama). CRITICAL WARNING: Costs can escalate rapidly, one documented case hit $5,000 processing 1.18B records with AI_COMPLETE. Cortex Analyst provides specialized text-to-SQL with claims of 90%+ accuracy using semantic modeling on your data. Cortex Search enables hybrid (vector + keyword) search on unstructured data. Cortex ML includes classification, forecasting, anomaly detection, and contribution analysis. Page should cover: function-by-function guide with examples, model selection recommendations, cost estimation and monitoring, Cortex Analyst setup and semantic modeling, Cortex Search configuration, comparison with native database functions, and cost management best practices.