OpenAI announces operator-grade GPT-5 for enterprise, $200/seat/month
OpenAI launched GPT-5 Enterprise with a new "operator mode" that reduces hallucinations by 40% on structured tasks. The model includes built-in tool use, persistent memory, and an audit log — positioning it squarely against Anthropic and Google for the corporate stack.
OpenAI announced GPT-5 Enterprise this morning, a tier of GPT-5 designed for regulated and operational workloads. Pricing starts at $200 per seat per month with annual contracts.
Operator Mode
The new operator mode constrains GPT-5 to a structured tool-use loop with explicit pre- and post-conditions. OpenAI says this reduces hallucinated tool calls by 40% on internal evals and produces a verifiable audit trail for each action taken.
Built-in Memory & Audit Log
Each enterprise seat ships with a managed long-term memory store, an immutable audit log, and granular RBAC. Memories can be inspected and revoked by admins; every model action is signed and time-stamped.
Competitive Position
The launch directly targets Anthropic's Claude for Work and Google's Gemini Enterprise. OpenAI is also bundling a fine-tuning quota and a dedicated SLA, making this the most aggressive enterprise push since the company's founding.