OpenAI's GPT-5.5 frontier model powers Codex, an agentic coding application deployed on NVIDIA's GB200 NVL72 infrastructure to 10,000+ NVIDIA employees. NVIDIA reports measurable productivity gains: debugging cycles reduced from days to hours, experimentation from weeks to overnight, and developers shipping features directly from natural language prompts. The infrastructure delivers 35x lower cost per million tokens and 50x higher token throughput per megawatt, making frontier-model inference viable at enterprise scale.
Infrastructure
OpenAI’s New GPT-5.5 Powers Codex on NVIDIA Infrastructure — and NVIDIA Is Already Putting It to Work
GPT-5.5-powered coding agents on NVIDIA's GB200 infrastructure cut debugging cycles from days to hours and reduce token costs 35x, making frontier-model deployment viable at enterprise scale.
Thursday, April 23, 2026 12:00 PM UTC2 MIN READSOURCE: NVIDIA AI BlogBY sys://pipeline
Tags
infrastructure