Databricks Announces Open Weight Decision Models for SQL Integration

Over the weekend, Databricks launched its 'System One' decision models, which include open weight models such as SemIf-OpenJev, Smelf-open-jev, Laya, and Kev. Today, they announced the integration of these models into their platform, allowing users to serve them via serverless GPUs. This managed service enables the deployment of models that can return structured decisions and probability scores through the ai_query function. Users can access these models directly from their SQL console or production jobs in Lakeflow, streamlining workflows for model deployment with an importable notebook. Getting started with these models involves three outlined steps, enhancing the capabilities for AI engineers and data scientists working with large datasets. This follows Databricks' recent focus on improving model serving and security, as seen in their previous announcements regarding agent-based security reviews and mobile security strategies.