# Databricks — Databricks partners with OpenAI on GPT-5.5

- Company: Databricks (databricks.com)
- Announced: 2026-04-23T23:00:00+00:00
- Category: partnership-acquisition
- Coverage: not counted
- Announcement: yes
- Group: announcements
- Source: https://www.databricks.com/blog/databricks-partners-openai-gpt-55
- Record: https://forck.live/items/25348-databricks-partners-with-openai-on-gpt-5-5
- Subject: Mosaic AI
- Models affected: GPT-5.5, GPT-5.4, Codex

Databricks is partnering with OpenAI to bring GPT-5.5, OpenAI's latest frontier model, to Databricks, where it will be governed through Unity AI Gateway. Databricks says GPT-5.5 is OpenAI's strongest frontier model for agentic work in enterprise, complex document reasoning and long-horizon coding agents, and that it now powers Codex, OpenAI's coding agent. On OfficeQA Pro with Oracle PDF and web search, GPT-5.5 scored 64.66%, up from GPT-5.4's 57.14%, which Databricks calls a new state of the art on the benchmark. In the full-agent OfficeQA Pro Agent Harness eval, where the model finds and parses documents itself using the Codex agent harness, GPT-5.5 scored 52.63% against GPT-5.4's 36.10%, a 46% reduction in errors. Databricks says the model can write and debug code, research online, analyze data, create documents and spreadsheets, and operate software across tools until a task is finished. GPT-5.5 and Codex are coming soon to Databricks.

## Evidence

Verbatim from https://www.databricks.com/blog/databricks-partners-openai-gpt-55:

> In a full-agent workflow eval (OfficeQA Pro Agent Harness), where the model must find the right documents, parse them, and compute answers on its own using the Codex agent harness, GPT-5.5 scored 52.63%, up from GPT-5.4's 36.10%. That's a 46% reduction in errors, showing that GPT-5.5's gains aren't just theoretical; they hold up in realistic, end-to-end enterprise workflows.

---

Record: https://forck.live/items/25348-databricks-partners-with-openai-on-gpt-5-5
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md