OpenAI has integrated its latest GPT-5.5 model into Codex, its agentic coding platform, marking a shift in how developers handle complex programming tasks. The upgrade runs on NVIDIA GB200 NVL72 rack-scale systems, enabling faster processing and deeper problem-solving capabilities. This deployment follows NVIDIA’s announcement that it has already begun using the system for internal development workflows.
Codex’s new foundation allows it to process larger datasets and generate more sophisticated code snippets compared to previous versions. OpenAI states the model improves accuracy in debugging and refactoring tasks, areas where earlier iterations struggled. The company has not disclosed user adoption numbers but confirms the transition is underway.
NVIDIA’s role extends beyond hosting the model. The company is optimizing the infrastructure to support real-time collaboration between AI agents and human developers. Engineers at NVIDIA report reduced latency in code generation tasks, a critical factor for teams working across global time zones.
The collaboration highlights a growing trend where AI tools are no longer confined to experimental phases but are being embedded into core business operations. OpenAI and NVIDIA have not yet announced pricing changes for Codex, which remains available through OpenAI’s API.
Source: blogs.nvidia.com