OpenAI introduced a websocket-based execution mode for its API, reducing latency in real-time agentic workflows. The upgrade allows agents to receive streaming function outputs directly over persistent connections instead of repeated HTTP polling.
For enterprise teams building autonomous agents at scale, the latency cut improves responsiveness in time-sensitive operations like live monitoring, fraud detection, and real-time customer support workflows.