Executive Summary
OpenAI has announced its latest frontier model, GPT-5.5, which powers its agentic coding application, Codex. The model runs exclusively on NVIDIA's new GB200 NVL72 systems, highlighting a deep, long-standing partnership between the two companies. As a massive early adopter, NVIDIA has already deployed the GPT-5.5 powered Codex to over 10,000 of its employees, reporting major productivity gains in a secure, enterprise-grade environment.
Key Takeaways
* New Model: OpenAI has launched GPT-5.5, its latest and most capable frontier model.
* New Application: The model powers Codex, an "agentic coding application" designed to assist developers and knowledge workers with complex problem-solving and innovation.
* Exclusive Hardware: GPT-5.5 runs on NVIDIA's GB200 NVL72 rack-scale systems, which reportedly offer 35x lower cost per million tokens and 50x higher token output per second per megawatt.
* Massive Internal Adoption: NVIDIA is the first major user, providing access to over 10,000 employees across functions like engineering, legal, marketing, and sales.
* Reported Gains: Internal NVIDIA teams report significant productivity improvements, reducing debugging cycles from days to hours and enabling end-to-end feature creation from natural language prompts.
* Enterprise Security: The deployment emphasizes security with a zero-data retention policy, with agents operating in sandboxed cloud VMs via remote SSH connections and using read-only permissions.
Strategic Importance
This announcement solidifies the symbiotic relationship between OpenAI's model leadership and NVIDIA's hardware dominance, while serving as a powerful enterprise case study for deploying agentic AI at scale.