Run OpenAI Models Directly on Databricks (www.databricks.com)

OpenAI and Databricks announced a partnership that brings OpenAI’s latest models, including GPT-5, natively to the Databricks Data Intelligence Platform so enterprises can run these models directly on their data with built-in security and governance. Organizations will be able to call GPT-5 from SQL, expose it as API endpoints via Model Serving, or embed it in Agent Bricks to build data-aware agents. Databricks layers in Mosaic AI Gateway for access control, monitoring and policy enforcement, and the integration will be available across AWS, Azure and GCP with a public preview via account teams. This matters because GPT-5 introduces stronger instruction-following and agentic tool use—reliably executing multi-step workflows, coordinating across tools, and adapting to evolving context—making it more suitable for end-to-end automation. Databricks adds a production-ready stack: Agent Bricks provides a built-in evaluation loop that measures accuracy on domain-specific tasks, tunes models, and iterates for higher quality; Model Serving and SQL access simplify deployment and enterprise-scale operations. The collaboration includes co-development for feedback loops and platform-specific optimizations, positioning customers (e.g., Experian, Mastercard) to build trusted, governed AI agents that deliver measurable quality and compliance on proprietary data.