OpenAI logged five separate incidents on 6 October 2026, touching login, ChatGPT conversations, image generation, and the Responses API. The longest ran 4 hours 28 minutes, and all five were OpenAI's own infrastructure, not anything in your integration.
Started
00:40 UTC
Duration
Lasted 57m
Source
IsDown
Next time
OpenAI is degraded. Is your product working?
An incident on their side does not always mean an outage on yours, and the only way to know is to already be watching your own endpoints. Set that up in under a minute and you will be notified the next time your own site is affected. And there will be a next time.
50 monitors for your own websites. No card.
Affected components: APIs, Responses, ChatGPT, Web, Image Generation, Login
What happened?
The day opened at 00:40 UTC with elevated errors in the API Platform's login and administration APIs, resolved in 57 minutes after OpenAI applied a mitigation. Just over an hour later, at 02:41 UTC, a second incident hit a wider set of products at once: ChatGPT conversations, image generation, Spaces and Pages, API rehydration, and Dots. That one closed in 37 minutes. Things were quiet until 13:35 UTC, when ChatGPT Go conversations started returning elevated error rates for just over an hour. The most disruptive incident of the day began at 15:45 UTC: errors processing large PDFs in the Responses API, which took 4 hours 28 minutes to clear. OpenAI's last public update on that one, at 16:39 UTC, said only that a mitigation was being implemented, with nothing further filed to confirm resolution details. A fifth incident, described by OpenAI simply as "Computer Use Is Experiencing Issues," started at 17:15 UTC and ran 2 hours 46 minutes affecting the agents API and computer use features, with no later update beyond the initial investigating note. Across all five incidents, OpenAI posted nine updates total, most following the same pattern of an investigating note followed by a short monitoring or identified note, then nothing more. No cause is given for any of the five. No user reports were filed in this window. OpenAI's own history shows this kind of clustering is normal for them: 107 incidents in the last 90 days, with a median duration of 1 hour 52 minutes, and 959 incidents all time, averaging 15 a month.
Learning
If you build on OpenAI's API, the lesson from a day like this is that failures are narrow and specific rather than one big blackout. Login broke separately from image generation, which broke separately from PDF processing in the Responses API, which broke separately from Computer Use. A health check against your own endpoints would not catch any of this, because the fault sits entirely on OpenAI's side of the line, in a particular feature path rather than the whole platform. UptimeRobot's own 30-day figures for OpenAI show 47 incidents, all resolved, with a median duration of 1h 21m, which lines up with the pattern you see here: short, frequent, feature-specific breaks rather than prolonged total outages. That means your own uptime dashboard can look perfectly green while a specific OpenAI feature your product depends on is down for hours. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.
