OpenAI's customer story records Third Moment's reported concurrency, duration, token use and supervision architecture; the operating results are not independently verified.
Third Moment reports running 10,000 agents for multi-day research
2026-10-14Verified announcementWATCH
An OpenAI customer story says Third Moment Research runs up to 10,000 concurrent agents for 36 to 48 hours, consuming roughly one trillion OpenAI tokens per day. Its default worker is Codex inside Docker, while outer GPT-6 Sol and GPT-6.1 Sol agents retry, terminate or escalate stalled work. The article is a vendor-hosted customer account and provides no raw telemetry or independent validation.
What this means for you
This is a concrete account of long-running agent operations at unusually high scale, including supervision rather than only model quality. It is useful as an architecture signal, but teams should not generalize the reported throughput or optimization gains without workload-level evidence.
Evidence & related material
Technical details & scores
| Reported Scale | up to 10,000 concurrent agents, 36-48 hour sessions and about one trillion OpenAI tokens per day |
|---|---|
| Worker | Codex in Docker |
| Supervisor | outer GPT-6 Sol and GPT-6.1 Sol agents retry, terminate or escalate stalled work |
| Reported Optimization | customer says Codex identified cache and memory packing that reduced one backtest from about three hours to 3-5 minutes |
| Evidence Limit | vendor-hosted customer story; no raw telemetry, cost breakdown or independent verification |
| Time Basis | OpenAI RSS published timestamp 2026-10-14T07:00:00Z |
Technical impact 3
Engineering value 4
Adoption signal 4
Maturity 3
Verification cost 4
Risk 4