Databricks is teaming up with OpenAI to bring the company's latest frontier model, GPT-5.5, to its enterprise data and AI platform. This collaboration aims to leverage GPT-5.5 for complex document reasoning and long-horizon coding tasks.
The partnership will see GPT-5.5 powering OpenAI's Codex, the coding agent, promising enhanced reasoning and execution capabilities for developers. This integration is positioned as a step towards a new paradigm of work where AI can handle more complex, multi-part tasks autonomously.
Databricks highlighted GPT-5.5's performance on its OfficeQA benchmark, designed to simulate enterprise workloads involving document analysis and multi-step calculations. On OfficeQA Pro, with PDF and web search retrieval, GPT-5.5 achieved a 64.66% score, a notable improvement over GPT-5.4's 57.14%.
A Leap in Enterprise Workflows
In more realistic end-to-end agentic workflows, where the model must independently find, parse, and compute answers, GPT-5.5 scored 52.63%. This represents a 46% reduction in errors compared to GPT-5.4, demonstrating tangible gains in practical enterprise scenarios. This advancement is crucial for AI model performance benchmarks, and you can read more about OpenAI's GPT-5.5: A Leap in AI Coding Assistance.
GPT-5.5 is expected to be available on Databricks soon, enabling businesses to utilize advanced reasoning capabilities directly on their own data, secured and scaled through Databricks' Unity AI Gateway.
