OpenAI's GPT-5.5, a powerful frontier model designed for agentic enterprise tasks, complex document reasoning, and long-horizon coding agents, is now natively supported on Databricks. This integration allows enterprise customers to apply GPT-5.5 to their proprietary data with end-to-end governance.
Customers can leverage GPT-5.5 to enhance coding workflows, whether with OpenAI's Codex or other coding agents, build sophisticated agents grounded in enterprise data, inject deeper intelligence into document processing pipelines, and integrate data insights seamlessly into daily operations.
All GPT-5.5 usage within Databricks is comprehensively governed through the Unity AI Gateway. This gateway provides customers with centralized security, granular cost controls, and robust observability from the outset, covering all agents, queries, and coding workflows.
The Unity AI Gateway acts as a unified governance layer for both GPT-5.5 model inference and Codex coding workflows, managed through a single control plane. It introduces enterprise-grade controls for GPT-5.5, including:
- Permissions and rate limits: Enabling access control per user and group to prevent excessive costs.
- Guardrails: Detecting Personally Identifiable Information (PII), blocking prompt injection attacks, and enforcing content safety policies, all configurable per endpoint.
- MCP governance: Auditing every agent tool call with full traceability.
- Automatic failover: Automatically routing traffic to backup models if rate limits are exceeded.
- Observability: Logging every request, including model calls and Codex interactions, with identity, tokens, latency, and cost data stored in user-owned Delta tables.
- Single bill: Consolidating billing across GPT-5.5, Codex, and other integrated models/coding tools.
Furthermore, Databricks enables users to deploy GPT-5.5 with full enterprise context. The platform's natural language analytics experience, Genie, which is powered by multiple frontier LLMs including GPT-5.5, allows business users to interact securely with complex enterprise data using plain English. This facilitates data exploration, ad-hoc query answering, and automation of knowledge work that traditionally required significant time. Users can also build custom agents using Agent Bricks, leveraging GPT-5.5's enhanced execution and reasoning capabilities.