Smart Routing in Unity AI Gateway: Match frontier quality with 30%+ lower cost per task
Summary
Smart Routing in Databricks Unity AI Gateway automatically directs coding tasks to the optimal model and harness across the price/performance frontier, matching frontier-level quality while cutting cost per task by more than 30%. The feature addresses the growing diversity of models and harnesses practitioners face by handling that selection for them rather than requiring manual tuning.
Summary generated by brickster.ai. For the full article, follow the source link above.
More from Databricks Blog
Introducing Unity Gateway Insights
Unity Gateway Insights now automatically uses Genie to analyze agent traces alongside your business data, pinpointing token waste from recurring MCP errors, simple prompts sent to expensive frontier models, and redundant workflows. Practitioners can generate tokenomics reports directly within AI Gateway, drill into linked session traces, and query Genie with custom questions to optimize their coding agent rollouts.
What's new for Genie in Microsoft: Copilot Cowork and Chat, Microsoft Teams, and Excel.
Databricks Genie is now generally available in Microsoft Teams, alongside expanded availability across Excel, Copilot Studio, and Microsoft Copilot Cowork. Users can query governed data estate insights directly in Teams via mentions and direct messages or connect Genie One to Copilot chat using the Azure Databricks Genie plugin.
How AUMOVIO centralized AI access with Databricks Unity Gateway
AUMOVIO consolidated fragmented LLM access across the enterprise by implementing Databricks Unity Gateway as a centralized, governed entry point for all AI workloads. Managed via Terraform, the architecture routes vendor-neutral model aliases across multiple foundation model providers while unifying access controls and chargeback tracking using Databricks system tables.
How we raised Databricks cluster rate limits by 7.5×
Databricks has raised default cluster-creation request-rate limits by 7.5× and eliminated over a hundred customer-specific overrides. The upgrade relies on adaptive concurrency limits that dynamically adjust to system health, paired with bounded queues and fair scheduling to absorb burst traffic.
Branch, sync, serve: What our Lakebase Hackathon taught us about building on Lakebase
Operational Postgres is becoming an integrated part of the lakehouse rather than a standalone system running alongside it. Customer hackathon projects demonstrate this shift by using Databricks Lakebase to safely test applications and AI agents via database branching, while syncing directly with the lakehouse to serve real-time data and ML features.
Azure Databricks sessions at Microsoft Ignite 2026
Azure Databricks is presenting eleven sessions at Microsoft Ignite 2026 focused on governed AI, AI agents, model fine-tuning, and Genie integrations across Microsoft Teams and Microsoft Copilot Cowork. Attendees can connect with the team November 17–20 in San Francisco at Booth #5271R in The Hub, the Ask the Experts kiosk, or The Social @ Canopy Hotel.