How Third Moment Research runs 10,000 agents for 36+ hours

Third Moment Research has demonstrated a large-scale agent deployment, running 10,000 agents simultaneously for more than 36 hours. The setup leverages a combination of models including GPT-6 Astra and GPT-6.1 Sol, alongside Codex and Daybreak, which are applied to cybersecurity tasks.

The key differentiator highlighted is that these agents maintain sustained focus over multi-day periods rather than operating in short bursts. They also reportedly improve with continued use, suggesting some form of adaptive or learning behavior during extended runs.

Scaling to 10,000 concurrent agents is notable because it pushes infrastructure and orchestration limits. Managing thousands of agents over extended durations raises questions about reliability, coordination, and resource efficiency that are central to practical agentic AI deployments.

The cybersecurity application context is relevant, as security operations often require persistent monitoring and response. Agents that can operate for days without degradation could reduce the need for human oversight in routine threat detection and triage workflows.

This case study illustrates a shift from single-agent demonstrations toward large-scale, long-running agent fleets, which may become a defining capability for AI labs and enterprises building autonomous systems.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.