OpenAI logged two separate incidents on 12 November 2026. One hit GPT-image-2.5 and 2.6 Sunburst and Flare with elevated errors for 35 minutes, and a second knocked out the Microsoft 365 connector for 29 hours 52 minutes. Both faults were OpenAI's to fix, not yours.
Started
08:52 UTC
Duration
Lasted 35m
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: Images, APIs
What happened?
OpenAI's status page recorded two incidents on this day, affecting the Images and APIs components. The first started at 08:52 UTC, when OpenAI said users were seeing elevated errors on GPT-image-2.5 Sunburst and Flare. Six minutes later it widened that statement to include 2.6 Sunburst and Flare as well, so the initial scope was incomplete at first posting. By 09:18 OpenAI said a mitigation had been applied and it was monitoring for recovery, and the incident closed out at 35 minutes total. The second incident started at 18:29 UTC and was described as an issue with Microsoft connectors, where new connections could break and tools would not work as expected. OpenAI gave no further detail on what was broken inside the connector at that point, only that it was investigating. The next update on record came at 16:20, reporting signs of recovery and continued monitoring, which puts the gap between updates at well over a day. That incident's logged duration is 29 hours 52 minutes, making it the longer of the two by a wide margin. No user reports were filed during either window on this record. OpenAI published five updates across the two incidents, three for the image error rate and two for the connector issue, and no post-mortem is on the record for either. For context, OpenAI logged 104 incidents in the preceding 90 days with a median duration of 1 hour 42 minutes, so the connector incident ran far longer than OpenAI's own typical fix time.
Learning
The Microsoft 365 connector incident is a good example of a failure that sits on OpenAI's side of a third-party integration, where your own product code and your own endpoints can be fully healthy while the connection between OpenAI and Microsoft's services is broken. A check that only pings your application or your API keys would show green the entire time, because the fault was not in anything you control or can restart. The same logic applies to the image model errors earlier in the day: elevated error rates on a specific model variant do not necessarily show up as a down signal on a generic health check, especially if your own calls happen to avoid the affected variant. Over OpenAI's last 30 days, 34 incidents were logged with a median duration of 1 hour 58 minutes, which is a reminder that short, model-specific error spikes are a recurring pattern worth watching for rather than a one-off. Your monitoring covers your half, UptimeRobot now watches the provider's half too, so next time you know which side broke without guessing.