How Third Moment Research runs 10,000 agents for 36+ hours
OpenAI's new models power agents that run for 36+ hours across 10,000 parallel instances
OpenAI showcases Third Moment Research running 10,000 agents continuously for 36+ hours using GPT-6 Astra, GPT-6.1 Sol, Codex, and the Daybreak cybersecurity model. This demonstrates a significant leap in long-horizon, large-scale agentic deployments that improve with use. It signals that sustained, massively parallel AI agent workflows are now production-viable on OpenAI's latest model family.