databricks Databricks Release Notes ·

Smart Routing: Cut AI costs by routing each task to the right model

aidatabrickspreviewdata-scientist
feature

Databricks introduces Smart Routing (Beta), an automated system that selects the optimal AI model and agent harness for each task. This feature routes work based on complexity, permissions, and budget, directing routine tasks to lower-cost models and reserving advanced models for more challenging problems. It aims to reduce overall AI task costs by over 30% while matching frontier-level quality, with all routing decisions visible and auditable. Smart Routing is compatible with Claude Code and Codex when used through Omnigent or ucode, and can route across proprietary and open models.

Features (1)
  • Smart Routing for Automated AI Model and Agent Selection

    Smart Routing (Beta) automatically picks the right AI model and agent harness for each task, considering factors like complexity, permissions, and budget. This system shifts routine work to lower-cost models while reserving frontier models for hard problems, with all routing decisions visible and auditable.

Read the original announcement →

https://docs.databricks.com/aws/en/release-notes/unity-gateway#smart-routing-cut-ai-costs-by-routing-each-task-to-the-right-model

Related releases