OFICIAL Snowflake News

Snowflake Decision Fast: Low-Latency AI Decision Models

What happened
Based on Snowflake News · Oct 19, 2026

Snowflake expands Snowflake Decision with lower-latency models and REST API support for real-time, structured AI decisions in workflows like fraud checks and agent triage.

Snowflake Decision Fast: Low-Latency AI Decision Models
Snowflake News — Snowflake
Key points
·
Snowflake introduced snowflake-decision-fast, a lower-latency variant of its decision model for high-volume, latency-sensitive tasks in private preview.
·
The model returns structured outputs like categories, scores, or yes/no probabilities with per-option probabilities for direct workflow integration.
·
Snowflake demonstrated real-time text classification with three simultaneous decisions—routing, urgency, and sentiment—returning results in under 300 ms.

Snowflake has introduced a lower-latency variant of its Snowflake Decision model, designed to handle high-volume, latency-sensitive tasks such as live routing, fraud detection, and agent triage. The new model, snowflake-decision-fast, prioritizes response speed without compromising decision quality, addressing customer demand for 'arcade speed' in interactive workflows. It is now available in private preview alongside expanded capabilities, including REST API support via Cortex AI Gateway and native SQL access through AI_COMPLETE.

The company demonstrated the model’s real-time capabilities with a text classification application that evaluates support tickets as users type. The application makes three simultaneous decisions—routing to the correct team, assessing urgency, and gauging customer sentiment—using structured outputs like categories, scores, or yes/no probabilities. Each decision is returned with per-option probabilities, enabling developers to set thresholds for live workflows, while construction bounds ensure only valid actions are returned, eliminating the need for error handling.

Snowflake also highlighted the model’s scalability for high-volume tasks, such as triaging millions of support tickets daily. By using snowflake-decision-fast in AI_COMPLETE, teams can apply classification and routing logic directly to large datasets via SQL, avoiding the complexity and cost of external API integrations. The model’s structured outputs are ready for downstream processing, making it suitable for offline scoring, asynchronous triage, and real-time REST API calls.

The company emphasized its commitment to improving both decision accuracy and execution speed, with ongoing updates planned to reduce latency and compute costs. Snowflake is iterating rapidly to push the frontier of structured AI decision-making, positioning snowflake-decision-fast as a solution for developers seeking precise performance profiles. Early access is available by contacting a Snowflake account team, with comprehensive private preview documentation to follow.

Original source → Deals on Clipraptor.com →