Snowflake Cortex
SQL-callable LLM functions and agents that run inside Snowflake
Overview
In-database AI for teams already on Snowflake: call LLMs, classify, extract, embed and search with plain SQL, plus Cortex Analyst for natural-language querying and Cortex Agents for multi-step workflows. Aimed at data engineers and analysts who don't want a separate AI stack.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 13 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Snowflake Cortex
- SQL functions: AI_COMPLETE, AI_CLASSIFY, AI_EXTRACT, AI_EMBED
- Cortex Analyst: natural-language-to-SQL over your tables
- Cortex Search: managed semantic/vector retrieval
- Cortex Agents and Snowflake Intelligence for multi-step analytics
- MCP server connectors (Salesforce Agentforce, UiPath, Anthropic)
- Microsoft Teams and Copilot integration (preview)
- Per-token AI Credit billing with no data leaving Snowflake
ASEAN Perspective
Snowflake Cortex in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
If your data already lives in Snowflake, Cortex spares you from standing up a separate AI pipeline: you call an LLM the same way you'd call any SQL function. AI_COMPLETE, AI_CLASSIFY and AI_EXTRACT cover the common jobs, Cortex Analyst turns plain-English questions into SQL, and Cortex Search handles vector retrieval. The catch is cost control. Since July 2026 the AI features bill in a separate "AI Credit" currency (roughly $2.00 per credit on global routing, $2.20 regional), and model choice matters enormously — Snowflake's own examples show a 59x spend gap running the same workload on different models. Analysts who stay on the platform all day will value not moving data out; teams doing high-volume inference need to watch token counts and pick smaller models on purpose. Skip it if you want provider flexibility or the lowest per-token rate on the market.
What people say
Almost everything written about Cortex in 2026 circles back to one subject: cost. The feature set rarely gets argued about — SQL-native LLM calls, semantic search and NL-to-SQL are genuinely useful — but the billing model is where teams get burned, and most published analysis is really a cost post-mortem.
The big change landed in July 2026, when Snowflake split AI usage into a dedicated, edition-independent "AI Credit." AI Functions, Cortex Search, Agents, Snowflake Intelligence and the REST API now bill at around $2.00 per credit on global routing and $2.20 on regional, with per-token LLM rates running roughly $0.12 to $5.10 per million tokens. Cortex Analyst, which converts natural language to SQL, comes bundled with Enterprise Edition and is metered separately at about 67 credits per 1,000 messages. Snowflake also cut AI_TRANSCRIBE pricing by up to 60% effective June 1, 2026.
The most-repeated warning is model selection. One documented case ran the same workload to the same business outcome on two different Cortex models and saw a 59x cost difference — a reminder that defaulting to the largest model quietly wrecks budgets. Reviewers consistently tell teams to benchmark small models first and monitor token spend from day one.
On the plus side, practitioners like that data never leaves the account — no separate vector DB, no export pipeline, no second vendor's security review. The 2026 roadmap added native Microsoft Teams and Copilot integration (in preview) and an MCP server that connects Cortex to Salesforce Agentforce, UiPath and Anthropic platforms, which widened its appeal beyond pure SQL shops.
What's thin is independent user sentiment. There's no large body of G2/Capterra-style Cortex-specific reviews yet; most coverage is vendor material or cost-engineering blogs rather than day-to-day operator feedback. The consensus you can extract is narrow but consistent: convenient if you're already committed to Snowflake, powerful for adding AI to existing pipelines, and unforgiving on cost if nobody owns model choice and token monitoring.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Snowflake Cortex's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Snowflake Cortex unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Snowflake Cortex directly →
Spotted something out of date? Suggest an update →
More in Code & Dev Tools
