The evidence
Caught by batchwatch’s own probes: a run of batch jobs that stopped completing on a model we watch directly.
Recovered after 28 min.
The record
| Provider | OpenAI |
|---|---|
| Model | gpt-5-nano |
| Started | 2026-09-10T23:46:04.000Z |
| Ended | 2026-09-11T00:14:25.000Z |
| Duration | 28 min |
| Measurements behind it | 2 |
This is a measurement of queue time — how long batch jobs actually waited — not a statement from OpenAI. batchwatch reports an outage only when a majority of established contributors, or our own probes, see the same slowdown at once. The full history is available as JSON and as an Atom feed.