OpenAI logged three separate incidents on 21 October 2026, touching ChatGPT, Codex, APIs and the FedRAMP-specific Astra and Sol services. The longest ran 5 hours 18 minutes, and all three were OpenAI's own infrastructure, not anything in your integration.
Started
02:26 UTC
Duration
Lasted 23m
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, Chat, Image Generation, Deep Research, Web, Login, Images, ChatGPT, Codex, Batch, Compliance, Realtime, Moderations, Fine-tuning, Embeddings, Voice mode, Operator, Search, File uploads, Files, GPTs, Audio
What happened?
The first incident started at 02:26 UTC, described by OpenAI as degraded performance in ChatGPT, Codex and APIs. It was short: a mitigation update followed at 02:27, and the incident closed out at 23 minutes total. The second and longest incident began with an identified status at 13:42 UTC, before the formal start time of 14:17, covering elevated FedRAMP API error rates for Astra and Sol. That one ran 5 hours 18 minutes, with monitoring confirmed at 16:46 UTC. The third incident started at 15:34 UTC, affecting gpt-5.6-luna API requests specifically, and was resolved within 1 hour 12 minutes after a mitigation update at 15:47. Across the day, OpenAI's status page lists 23 components as potentially affected, though the three incidents named only a subset: ChatGPT, Codex, APIs, and the FedRAMP-specific Astra and Sol services, plus gpt-5.6-luna requests. OpenAI published seven updates total across the three incidents, moving each through investigating or identified to monitoring, with no further detail on root cause in any of them. No user reports were filed during the window on this page. For context, OpenAI has logged 102 incidents in the past 90 days, with a median duration of 1 hour 38 minutes, so the middle incident here ran well above that typical length.
Learning
If you run a product on OpenAI's API, this is the layer you cannot see from your own side: the FedRAMP-specific Astra and Sol error spike lasted over five hours, but your own endpoint checks would have shown green the whole time because the fault sat inside OpenAI's infrastructure, not yours. UptimeRobot's own 30-day figures show 41 incidents against OpenAI, all resolved, with a median duration of 1h 36m, which lines up closely with OpenAI's own 90-day median of 1 hour 38 minutes. That consistency matters: it means the pattern on this page is typical, not an anomaly, and errors clustering around specific model names like gpt-5.6-luna or specific compliance tiers like FedRAMP are a recognisable shape of OpenAI outage, not 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.