OpenAI's launch-day outage hit ChatGPT, Codex, and the API for over 5 hours

OpenAI confirmed elevated errors across ChatGPT, Codex, and the API — including the Agents API — for 5 hours 21 minutes on its DevDay 2026 launch day, then marked the incident resolved.

#What happened

• OpenAI's status page first flagged "elevated errors affecting ChatGPT, Codex, and the API" at 17:52 UTC on September 29, 2026 (10:52 AM PT / 1:52 PM ET). Reported symptoms: failed requests, difficulty logging in or signing up, and tasks that did not complete.
• At 18:24 UTC the scope was widened to explicitly include the Agents API.
• At 19:39 UTC OpenAI applied a mitigation and moved to monitoring recovery, posting repeated "monitoring the recovery" updates at 21:09, 22:03, and 22:47 UTC.
• At 23:14 UTC the incident was marked resolved: "The issue causing elevated errors across ChatGPT, Codex, and the API has been resolved, and all affected services have recovered." OpenAI promised a detailed root-cause analysis within the next 5 business days.
• This is a separate incident from the later "Elevated errors in ChatGPT Space Pages" (first flagged at 02:20 UTC on September 30) already reported on dotsfeed.

#Why it matters

• The outage struck on the biggest day in OpenAI's product calendar — DevDay 2026, when it launched Dots, the Agents API public beta, GPT-6.1 Sol, and more than 20 other announcements.
• It was the Agents API's first public stress test, one day after its beta debut, and it failed for part of the afternoon.
• A promised RCA within 5 business days is the open question for developers weighing the API for production agent workloads.

Verified Sep 30, 2026.