ChatGPT API slowdown on 5 October 2026
ChatGPT API from San Francisco: slower than usual; 2026-10-05 15:00 UTC to 2026-10-05 16:30 UTC: peak first visible token 1.83 s vs a usual 1.1 s.
- What our checks saw
- Slower than usual
- Cities affected
- San Francisco
- When
- to
- How long
- 90 minutes
- Slowest check
- 1.83 s, usually 1.10 s
- OpenAI's status page
- Open it
Where our checks saw it
Each glowing city had slow or failing checks during the incident, shown at its worst. We test from 6 cities, so other places may have been hit too.
Our checks, city by city
Each point is one check every 30 minutes, from three hours before the incident to three hours after. Coloured points are the checks we counted as slower or failing. The shaded area shows how far above the usual time they were. Not affected: Amsterdam, Montreal, Singapore, Tokyo, Mumbai.
San Francisco
Every check as a table
| Time (UTC) | Check | Reading |
|---|---|---|
| 12:00 | 1.58 s | Normal |
| 12:30 | 1.20 s | Normal |
| 13:00 | 1.10 s | Normal |
| 13:30 | 1.68 s | Slower than usual |
| 14:00 | 1.82 s | Slower than usual |
| 14:30 | 1.72 s | Slower than usual |
| 15:00 | 1.94 s | Slower than usual |
| 15:30 | 1.23 s | Normal |
| 16:00 | 1.49 s | Normal |
| 16:30 | 1.56 s | Normal |
| 17:00 | 1.06 s | Normal |
| 17:30 | 1.08 s | Normal |
| 18:00 | 1.08 s | Normal |
| 18:30 | 1.63 s | Normal |
| 19:00 | 1.95 s | Slower than usual |
| 19:30 | 4.03 s | Much slower, or failing |
How fast is your API around the world?
Run a free speed test from multiple cities and find out where your users are waiting. No setup, no account required.
Test my API
No account required · Takes about 30 seconds.