Agents
OpenAI Codex Runs GPT-5.5 on NVIDIA Infrastructure, Powering Agent Workloads at Scale
NVIDIA published details on its infrastructure partnership powering OpenAI's Codex with GPT-5.5, enabling the agentic coding platform to handle complex multi-step agent workloads. GPT-5.5 matches GPT-5.4's per-token latency in real-world serving despite significantly higher intelligence, often reaching higher-quality outputs with fewer tokens and fewer retries. The premium GPT-5.5 Pro tier ($30/$180 per 1M input/output tokens) targets maximum accuracy for enterprise agent deployments requiring the highest reliability.
Source
↳ Follow the thread