OpenAI API Uptime History
Measured by PulsAPI's own checks, not reported by OpenAI API. Every figure below is computed over the 95 days in the last 365 that PulsAPI actually observed, from 2026-05-22 to 2026-08-24. Days nobody watched are left out rather than counted as perfect.
OpenAI API Uptime at a Glance
- Measured uptime
- 77.73%
- Days with an issue
- 60
- Longest outage
- 11d
- Mean recovery
- 9h 59m
across 95 observed days
days were not fully operational
single longest resolved incident
average time to resolve
Monthly OpenAI API Uptime
Each month is the mean of its observed days. The strongest month was August 2026 at 91.21%, the weakest July 2026 at 58.88%. Months with few observed days are marked, because a short month is not evidence of a good one.
| Month | Uptime | Uptime bar | Days observed |
|---|---|---|---|
| August 2026 | 91.21% | 24 | |
| July 2026 | 58.88% | 31 | |
| June 2026 | 88.90% | 30 | |
| May 2026 | 70.30% | 10partial |
How This OpenAI API Uptime Figure Is Measured
PulsAPI checks OpenAI API every 60 seconds and records the result. A day's uptime is the share of that day's checks that came back fully operational, so a 20-minute outage shows up as a dip rather than being rounded away. The window figure is the mean of those daily values.
Two things this deliberately is not. It is not OpenAI API's own published uptime, because vendors generally measure against their own SLA definitions, which usually exclude scheduled maintenance and often exclude partial degradation. And it is not a promise about the future: 95 days of history describes what happened, and the useful way to read it is alongside the incidents behind the dips.
Uptime measured this way is a mean of component uptimes where a vendor publishes components, so coverage matters as much as the percentage. That is why the observed-day count sits beside every figure on this page rather than in a footnote.
These measurements are published under CC BY 4.0. Quote or republish them freely, with attribution to PulsAPI.