StatOSS

OpenAI API uptime

OpenAI API, checked by StatOSS every 5 minutes from Helsinki in Finland, Falkenstein in Germany, Copenhagen in Denmark, Ashburn in the USA, Singapore and Sydney in Australia. Measured from outside, with every outage dated in UTC.

api.openai.com/v1/modelsEvery 5 minutes

OpenAI API is up.

Up for 51 d 10 h. Last checked 2026-11-20 18:53 UTC from Ashburn, USA: 401 in 71 ms.

Last 24 hours

Up for 51 d 10 h100% passed

287 checks, median 174 ms from Copenhagen, all passed

95% under 217 ms

Response time by region
EuropeEurope173 ms
N. AmericaNorth America93 ms
AsiaAsia253 ms
OceaniaOceania206 ms

Medians over this range, on one scale up to 300 ms. Europe's places are scaled to Copenhagen, its fastest shared place.

Last 90 days

Up for 51 d 10 h100% passed

14,804 checks, median 188 ms from Copenhagen, all passed

95% under 247 ms

Response time by region
EuropeEurope188 ms
N. AmericaNorth America92 ms
AsiaAsia254 ms
OceaniaOceania216 ms

Medians over this range, on one scale up to 300 ms. Europe's places are scaled to Copenhagen, its fastest shared place.

Each bar of the first strip is one check, five minutes apart; each bar of the second is a day. Height is response time, red is a failed check, amber a timeout. Times are UTC, as of 2026-11-20 18:56 UTC.

In the last 30 days OpenAI API passed 8,863 of 8,863 checks, 100%.

No outage since checks began on 30 Sep 2026.

Availability

The share of checks that passed, and the outages that began, in each window up to now.

LastPassedChecksOutages
24 hours100%288 of 2880
7 days100%2,014 of 2,0140
30 days100%8,863 of 8,8630
90 daysNot enough history yet: checks began on 30 Sep 2026

A year of checks is kept; the year's figure appears here once there is a year of them.

Response time by location

Over the last 7 days, from the checks that passed: the time from the start of the request to the answer's headers, with the DNS lookup, the connection and TLS included.

FromMedian95th percentileMeanChecks
Falkenstein, Germany204 ms263 ms204 ms42
Helsinki, Finland196 ms224 ms195 ms69
Ashburn, USA82 ms126 ms85 ms504
Copenhagen, Denmark181 ms220 ms182 ms393
Singapore250 ms298 ms254 ms503
Sydney, Australia212 ms258 ms216 ms504

Outages

None detected since checks began on 30 Sep 2026.

What is checked

The API behind GPT models, used by apps that call a language model, generate images or transcribe audio.

StatOSS sends GET https://api.openai.com/v1/models and counts the check as passed on status 401 within 10 seconds. The API's list of models, on api.openai.com, where requests to the models go too. Without a key it answers 401.

An answer to an unauthenticated request says the API is reachable, not that models answer or how fast. ChatGPT is a separate product.

OpenAI API reports its own incidents at status.openai.com. The figures here are measured from outside and can differ from it in both directions.

How it is measured

  • One check every 5 minutes, from one location at a time in turn: Helsinki in Finland, Falkenstein in Germany, Copenhagen in Denmark, Ashburn in the USA, Singapore and Sydney in Australia.
  • A check fails when no answer comes within 10 seconds, when the connection or TLS fails, or when the answer is not status 401.
  • A failed check is repeated at once from a second location, and counts only when that location fails too, or cannot answer within 20 seconds.
  • Two failed checks in a row make an outage, dated from the second, and the first check that passes ends it. A failure shorter than 10 minutes may not make one, and times are good to about 5 minutes.
  • Availability is the share of checks that passed. There are no maintenance windows: every check counts.

It is the same check a StatOSS status page runs for its owner, with the same rules; the docs on checks have the details.