NVIDIA has announced that OpenAI's latest large language model, GPT-5.5, now powers its Codex platform, running on NVIDIA's cutting-edge GB200 NVL72 rack-scale systems. This integration marks a significant leap in enterprise AI capabilities, enabling Codex to function as a sophisticated AI agent for complex software development tasks.
Codex, the AI-powered coding assistant, leverages GPT-5.5's advanced reasoning to move beyond simple code suggestions. It now handles end-to-end feature development, automated debugging, and even overnight experimentation—all triggered by natural language prompts. Developers can describe a feature in plain English, and Codex will generate, test, and refine the code autonomously.
More than 10,000 NVIDIA employees have already received early access to the GPT-5.5-powered Codex. According to NVIDIA, internal teams have reported dramatically faster iteration cycles, with some bugs fixed in minutes instead of hours. The system's ability to run around the clock on NVIDIA's high-performance infrastructure allows for continuous integration and testing.
The deployment utilizes the GB200 NVL72, a rack-scale system designed specifically for massive AI workloads. This hardware provides the computational muscle needed to run GPT-5.5's inference at scale, supporting multiple concurrent AI agents. The combination of OpenAI's model and NVIDIA's infrastructure positions Codex as a cornerstone for enterprise knowledge work, transforming how organizations approach software development.