OpenAI Launches GPT-5.5 — 'Smartest Model Yet' Powers New Codex Super-App on NVIDIA GB200 Infrastructure

OpenAI Blog·high signal

OpenAI released GPT-5.5 on April 23, its latest frontier model co-designed for and served on NVIDIA GB200/GB300 NVL72 rack-scale systems delivering 35x lower cost per million tokens. The model is described as faster, more token-efficient, and better at tool use than GPT-5.4 — it 'understands the task earlier, asks for less guidance, uses tools more effectively, and checks its work.' OpenAI committed to deploying 10+ GW of NVIDIA systems and launched next-gen Codex to paid ChatGPT tiers simultaneously.