Snowflake Cortex AI
by Snowflake Inc.
Turn conversations, documents and images into intelligent insights with AI next to your data.
Snowflake Cortex AI is a fully managed suite of generative-AI and LLM features built directly into the Snowflake Data Cloud. It lets teams run industry-leading large language models, build agents, and analyze multimodal data using SQL or APIs without moving data outside Snowflake's governed perimeter.
Snowflake Cortex AI is the built-in AI and LLM layer of the Snowflake Data Cloud, giving organizations serverless access to industry-leading models from Anthropic (Claude), Meta (Llama), Mistral and others directly where their data already lives. It comprises a family of capabilities including Cortex AI Functions (AISQL) for invoking LLMs through SQL to summarize, classify, translate, extract and embed text and multimodal data; Cortex Analyst, an LLM-powered text-to-SQL service that answers business questions over structured data via a REST API; Cortex Search, a low-latency hybrid retrieval engine that powers enterprise search and RAG; Cortex Agents, which orchestrate across structured and unstructured data using tools, planning and reflection; plus Cortex Fine-tuning, AI Guardrails, AI Observability and agentic coding assistants. Because everything runs inside Snowflake's security and governance boundary on serverless infrastructure, customers avoid moving sensitive data to external model providers, and Snowflake does not train its hosted models on customer data. Cortex AI is consumed on a usage-based, token-and-credit pricing model and is positioned as the AI engine for analytics, RAG and agentic applications for enterprises already standardized on Snowflake.
At a Glance
- Category
- Data & Analytics
- Pricing
- Usage-based
- Target Market
- CIOs, CTOs, Data Scientists, Data Engineers, Enterprise Developers, Data & Analytics Leaders
- Founded
- 2012
- Headquarters
- Bozeman, Montana, USA
- Customers
- Over 65% of Snowflake customers reported using Snowflake AI features such as Cortex Analyst and Cortex Agents (per Snowflake blog, 2025)
Key Features
- ✓Cortex AI Functions (AISQL)
SQL and Python functions that call hosted LLMs to summarize, classify, translate, extract, redact PII and embed text, images and audio directly in queries.
- ✓Cortex Analyst
A fully managed, LLM-powered text-to-SQL service that lets business users ask questions in natural language and get answers over structured data via a REST API.
- ✓Cortex Search
A low-latency, high-quality hybrid (semantic and keyword) search service that powers enterprise search experiences and retrieval-augmented generation.
- ✓Cortex Agents
Schema-level agent objects that orchestrate tools such as Cortex Analyst, Cortex Search and code execution using an LLM-driven plan-act-reflect loop over governed data.
- ✓Industry-leading hosted LLMs
Serverless access to models including Anthropic Claude, Meta Llama and Mistral Large, all running inside Snowflake's secure service perimeter.
- ✓Cortex Fine-tuning & AI Guardrails
Customize hosted models on proprietary data and apply built-in safety guardrails, observability and governance controls.
Capabilities
Use Cases
- •Unstructured data analytics
Summarize, classify and extract insights from documents, support tickets and other text or multimodal data using AISQL functions.
- •Natural-language business intelligence
Let non-technical users query structured data conversationally through Cortex Analyst without writing SQL.
- •Enterprise RAG and AI agents
Combine Cortex Search and Cortex Agents to build governed chatbots and agentic apps grounded in a company's own Snowflake data.
Ideal For
Best For
- ✓Running LLMs over enterprise data in SQL without data movement
- ✓Natural-language analytics and text-to-SQL for business users
- ✓Building governed RAG and agentic applications on the data platform
Integrations
Market & Ratings
Over 65% of Snowflake customers reported using Snowflake AI features such as Cortex Analyst and Cortex Agents (per Snowflake blog, 2025)
Market Analysis
Pros
- ✓No data movement keeps sensitive data inside Snowflake's governance boundary
- ✓Access to multiple top-tier LLMs through one managed service
- ✓Simple SQL-based invocation lowers the barrier for data teams
- ✓Fully managed serverless infrastructure removes ops overhead
Cons
- ✗Token-and-credit usage pricing can be unpredictable and expensive at scale
- ✗Tied to the Snowflake platform, limiting use outside it
- ✗Per-credit costs vary by edition and can exceed calling model providers directly
- ✗Some advanced features are in preview rather than general availability
Pricing
Usage-based (Cortex AI)
Usage-based (Snowflake credits / per-token)
- ✓Pay per token consumed by LLM and AISQL functions
- ✓Serving compute billed for managed AI services
- ✓Standard, Enterprise and Business Critical credit pricing tiers
- ✓Included with Snowflake account; new accounts get free trial credits
Cortex AI is billed on consumption: generative functions charge for input and output tokens (totaled, divided by 1M and multiplied by a per-credit rate), plus serving compute for services like Cortex Search and Analyst. Per-credit cost depends on the Snowflake edition (e.g., Standard vs Enterprise) and region. Snowflake lowered Cortex AI per-credit prices in April 2025. Costs can be high for large-scale jobs, so monitoring via Account Usage views is recommended.
Sources
This page was written from 8 sources, 7 on domains other than snowflake.com.
- 1.snowflake.com — cortexvendor
- 2.docs.snowflake.com — guides overview ai features
- 3.docs.snowflake.com — cortex analyst
- 4.docs.snowflake.com — cortex agents
- 5.docs.snowflake.com — cortex search overview
- 6.en.wikipedia.org — Snowflake Inc.
- 7.select.dev — snowflake cortex ai sql overview and cost monitoring
- 8.improvado.io — snowflake competitors and alternatives
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Quantum Metric Felix Agentic
Agents that watch your digital funnel, find what broke and price the damage
Zilliz Vector Lakebase
Unified vector lakebase merging real-time vector search, analytics and data lake queries on one copy of data
LanceDB
The multimodal lakehouse for AI — vector, full-text, and hybrid search at billion-row scale
Akeneo Agentic Ziggy
Agentic AI layer inside the Akeneo Product Cloud that enriches, governs, and orchestrates product data at catalog scale