Is OpenAI API down?

What works right now

• ●UpAPI front doorYes: api.openai.com is answering requests.401 in 236 ms

Failing for you? Check whether it is you

1. Is it you?

   401 means a bad or missing key. 429 is a rate limit, or insufficient_quota when the account is out of credit. 500, 502, 503 and timeouts are on their side. Check your tier's limits on our model reference.

What to do while OpenAI API is down

When OpenAI API is down, requests to api.openai.com fail or hang. The official page lists which parts (Chat Completions, Responses, Batch, Realtime...) are affected; ChatGPT incidents often do not touch the API.

1. Retry the right errors

   Retry 429 and 5xx with exponential backoff and jitter; never retry 400 or 401. Set a request timeout so a hung stream fails fast instead of holding a worker.

2. Fail over to another model

   Keep a second provider behind the same interface. Our price table finds a model at a similar price, and the model reference checks it has the context window, tool calling and structured output you rely on.

3. Queue, do not drop

   Write failed requests to a queue and replay them when the status page clears. Work that can wait a day is cheaper through the Batch API anyway.

Last 24 hours

Last 30 days

Incident timeline

Outages our checks saw, and incidents from the official status page, newest first.

1. OfficialOct 23, 21:30 UTC · 33 minHelp Center Chat Support Degraded

   All impacted services have now fully recovered.

2. OfficialOct 23, 3:41 UTC · 2.3 hDelays in customer support responses

   All impacted services have now fully recovered.

3. OfficialOct 22, 19:36 UTC · 4.5 hNew Ad Accounts failing to provision admins

   All impacted services have now fully recovered.

4. OfficialOct 22, 18:51 UTC · 2.1 hDegraded Performance of Dots on Enterprise

   All impacted services have now fully recovered.

5. OfficialOct 22, 2:18 UTC · 6.5 hElevated ChatGPT conversation and Responses API errors

   All impacted services have now fully recovered.

6. OfficialOct 21, 22:36 UTC · 1 minAuthentication errors for some users

   On October 20, from 9:35 to 10:35 a.m. PDT, some users near Illinois experienced authentication errors with the API, ChatGPT, and Codex. We rolled back a configuration change, and service was restored.

7. OfficialOct 21, 15:34 UTC · 1.2 hElevated error rates for gpt-5.6-luna API requests

   All impacted services have now fully recovered.

8. OfficialOct 21, 14:17 UTC · 5.3 hElevated FedRAMP API error rates for Astra and Sol

   All impacted services have now fully recovered.

9. OfficialOct 19, 13:31 UTC · 42 minFile Download errors

   All impacted services have now fully recovered.

10. OfficialOct 15, 16:29 UTC · 2.4 hWidespread feature degradation

    All impacted services have now fully recovered.

11. OfficialOct 14, 17:00 UTC · 41.7 hDelayed Costs data in the Compliance API

    All impacted services have now fully recovered.

12. OfficialOct 14, 10:18 UTC · 40 minLimited availability on gpt-image-2.5-flare and gpt-image-2.5-sunburst

    All impacted services have now fully recovered.

13. OfficialOct 9, 21:48 UTC · 1.6 hWorkspace admins may be unable to manage plugins

    All impacted services have now fully recovered.

14. OfficialOct 9, 17:23 UTC · 3 daysDelayed Costs data in the Compliance API

    All impacted services have now fully recovered.

15. OfficialOct 7, 23:42 UTC · 21 minElevated errors with GPT-5.6 Instant in FedRAMP workspaces

    All impacted services have now fully recovered.

16. OfficialOct 7, 1:12 UTC · 1.8 hWorkspace Agents Degraded and Responses Impacted

    All impacted services have now fully recovered.

Official status

The official page: https://status.openai.com/.

Get alerts, or show it in your README

Follow the RSS feed of OpenAI API incidents in any reader, or paste this badge into a README; it updates itself.

[![OpenAI API status](https://tensorust-status.vercel.app/api/badge?id=openai-api)](https://tensorust.com/status/openai-api/)

How we check OpenAI API

Every 5 minutes, from Washington, D.C., with no API key. A failed check is retried once after 2 seconds before it counts; OpenAI API is down when every critical check fails, and degraded when some do or the official page reports a problem.

• https://api.openai.com/v1/models: api.openai.com answering (401 without a key is the healthy reply).

Other services