ChatGPT, Codex And The API Broke For Five Hours.
OpenAI's status page listed 30 affected components on September 29. By the next morning it had logged three more incidents.
Server racks in a data center, shown for illustration. Photo: BalticServers.com, CC BY-SA 3.0, via Wikimedia Commons.
If ChatGPT kept failing on you Tuesday, it wasn't your connection. OpenAI's own status page says so.
At 10:52 AM Pacific on September 29, OpenAI opened an incident titled "Elevated errors across ChatGPT, Codex, and the API including the Agents API." It wasn't marked resolved until 4:14 PM Pacific, 5 hours and 22 minutes later.
How big it was
The incident page lists 30 affected components. On the ChatGPT side that includes Conversations, Login, File uploads, Voice mode, Search, GPTs, Image Generation, Deep Research, Agent and Connectors. For Codex it lists Web, CLI, the VS Code extension and the Codex API. The API list runs from Chat Completions and Responses to Embeddings, Batch, Realtime, Fine tuning and Files.
Android Authority reported that OpenAI told users to expect "failed requests, difficulty logging in or signing up, and tasks that do not complete." OpenAI said it applied a mitigation at 12:39 PM Pacific, then kept posting the same monitoring message for another 3 hours and 35 minutes before calling it resolved.
And then support went down
At 4:18 PM Pacific the status page resolved a second incident titled "Support available via email." It says the issue affecting support chat had been fixed and users could reach support through the Help Center again. So for part of Tuesday, the product was failing and the chat you'd use to report it wasn't working either.
Wednesday didn't start clean
By Wednesday morning, September 30, the status page had three more incidents: elevated error rates for ChatGPT Pro and Plus users, resolved at 6:23 AM Pacific; elevated errors in ChatGPT Space pages, under monitoring at 8:00 AM; and elevated latency for some API requests on Chat Completions and Responses, still being worked on at 10:13 AM. Plus and Pro are the paid tiers.
What OpenAI has promised
The resolution note says "The detailed Root Cause Analysis (RCA) will be published in the next 5 business days." We'll read it when it's out. Until then there's no public explanation of what failed.
Why this belongs here
This lands right after a stretch we covered last week, when the status page logged 12 incidents in 10 days. Developers who build on the API and teams that run Codex all day don't get their afternoon back when the status page turns green. Five hours is most of a workday.
Sources: OpenAI status page, September 29 incident, OpenAI status history, Android Authority.