Databricks integrates GPT-5.5 into workflows

OpenAI and Databricks said in May 2026 that GPT-5.5 is now available inside Databricks for enterprise agent workflows, extending the model from chat-style use into software that can operate inside governed business systems. (openai.com) Databricks described the rollout as native availability on its Data Intelligence Platform, with access through model serving, agent-building tools and Unity AI Gateway. OpenAI said Databricks is using GPT-5.5 for enterprise agent workflows after the model set a state-of-the-art result on the OfficeQA Pro benchmark. (docs.databricks.com) The companies framed the partnership around business tasks that require long context, tool use and action across multiple systems. ### Where exactly is GPT-5.5 showing up inside Databricks? (databricks.com) Databricks said GPT-5.5 is natively available on the platform and can be used to power agent building, natural-language data access through Genie, and governed model access through Unity AI Gateway. The company said the gateway adds permissions, cost controls, guardrails and observability for GPT-5.5 and Codex from the start. Databricks documentation lists a supported endpoint named `databricks-gpt-5-5` in its Foundation Model APIs. The documentation says the model accepts text and image inputs and supports a 400,000-token total context window with up to 128,000 output tokens. (openai.com) ### What are these “enterprise agent workflows” supposed to do? OpenAI said GPT-5.5 is designed for “complex tasks like coding, research, and data analysis across tools,” and said the model can plan, use tools, check its work and continue through multi-step assignments. Its API guidance says GPT-5.5 is suited to tool-heavy agents, grounded assistants, long-context retrieval and production workflows where execution quality matters. (databricks.com) Databricks said the model is aimed at enterprise agent workflows, complex document reasoning and long-horizon coding agents. In a separate Databricks post about running OpenAI models directly on its platform, the company said customers can call models within SQL, through model serving endpoints or via Agent Bricks to build domain-specific agents on enterprise data. (docs.databricks.com) ### Why does governance keep appearing in both companies’ descriptions? Unity AI Gateway is central to Databricks’ pitch. Databricks said the gateway governs GPT-5.5 and Codex with permissions, cost controls, guardrails and full observability, language that places the model inside existing enterprise controls rather than as a standalone assistant. (openai.com) OpenAI’s API documentation also points to operational constraints that matter in enterprise deployments. The company says GPT-5.5 has pricing tiers by token volume and adds a surcharge for regional processing endpoints tied to data residency. Those details matter when companies decide where to run models and how to monitor usage inside regulated or cost-sensitive environments. (databricks.com) That last point is an inference based on OpenAI’s pricing and regional-processing documentation. ### What evidence did the companies give for using GPT-5.5 here? OpenAI said Databricks adopted GPT-5.5 for enterprise agent workflows after the model posted a new state-of-the-art result on OfficeQA Pro. The OpenAI page also quotes Arnav Singhvi, a research engineer, saying, “GPT-5.5 has been great in terms of knowledge lift” and calling it “a step size function change” for knowledge work. (databricks.com) OpenAI said more broadly that GPT-5.5 improves on earlier models in agentic coding, computer use, knowledge work and scientific research, while matching GPT-5.4 per-token latency in serving. Databricks separately described GPT-5.5 as OpenAI’s strongest frontier model for agentic work in enterprise settings. (developers.openai.com) ### What happens next for customers using the platform? Databricks says GPT-5.5 and Codex are already available through its platform tooling, including Unity AI Gateway and Foundation Model APIs. OpenAI said on April 24, 2026 that GPT-5.5 and GPT-5.5 Pro became available in the API, giving Databricks customers a current production model rather than a preview release. (openai.com) OpenAI’s model page says GPT-5.5 is priced at $5 per 1 million input tokens and $30 per 1 million output tokens in the API, with higher charges for very long prompts and a 10% uplift for regional processing. Databricks’ next step for customers is operational rather than ceremonial: the model is already listed in supported-model documentation and can be deployed through the company’s managed interfaces. (openai.com) (developers.openai.com) (databricks.com)