OpenAI API outage — what to do when the API is down

When your production app returns HTTP 500 errors or times out completely, you need to know within 30 seconds whether OpenAI's API is down or your code broke. Check the [live OpenAI status](/status) on this site first — i

OpenAI API outage — what to do when the API is down

When your production app returns HTTP 500 errors or times out completely, you need to know within 30 seconds whether OpenAI's API is down or your code broke. Check the live OpenAI status on this site first — it updates every 60 seconds from OpenAI's official feed and shows component-level status for the API, ChatGPT, and Sora.

How to check if OpenAI is actually down

Start with the live status monitor on this page. It pulls directly from OpenAI's infrastructure and tells you instantly if the API is operational, degraded, or offline. If you see "Major Outage" next to "API" with a red indicator, it's not your code.

Head to status.openai.com for OpenAI's official status dashboard. Look at the timeline at the bottom — if there's an active incident tagged "API", you'll see when it started and updates from OpenAI's engineering team. The component breakdown shows whether it's the completions endpoint, embeddings, or specific models like GPT-4 or o1 that are affected.

Third-party monitoring helps confirm. Search "OpenAI API down" on Twitter/X in the last hour. Check downdetector.com/status/openai for spike patterns in user reports. If hundreds of developers are reporting 503 errors simultaneously, you're looking at an outage, not a local issue.

Is it OpenAI or is it you

Before you assume OpenAI is down, eliminate your own infrastructure. These steps take five minutes:

  • Test the example endpoint. Hit `https://api.openai.com/v1/models` with a simple GET request using curl or Postman. If this returns your model list, the API is reachable and your API key works.
  • Check your error codes. HTTP 500, 502, 503 indicate server-side problems at OpenAI. HTTP 429 means rate limits — that's you, not them. Read more about API error 429 and other API errors. HTTP 401 or 403 means authentication failure.
  • Isolate the model. If you're calling `gpt-4o`, try `gpt-3.5-turbo` instead. If GPT-3.5 works but GPT-4 times out, it's a specific model issue, not a full API outage.
  • Verify your quota. Log into platform.openai.com/settings/organization/billing. If you've hit your usage limit, requests fail silently or return errors. Add credits and retry.
  • Test from a different network. Run the same request from your phone's hotspot, not your office Wi-Fi. Some corporate firewalls block OpenAI's IP ranges.
  • Check latency, not just errors. Use `time curl` to measure response time. If requests succeed but take 45 seconds instead of 2 seconds, OpenAI's infrastructure is degraded even if it's technically "up".

What to do during an outage

When the API is confirmed down, switching into damage control mode protects your users and your budget:

  • Implement exponential backoff immediately. Don't retry failed requests every second — you'll burn through your quota once the API recovers. Use retry logic like: 1 second, 2 seconds, 4 seconds, 8 seconds, then give up. Most libraries have this built-in.
  • Queue requests instead of dropping them. If you're running a production service, use a message queue like Redis or RabbitMQ to hold incoming requests. Process the queue when the API comes back online. Your users see "processing" instead of "error".
  • Switch to cached responses if possible. For non-critical features, serve the last known good response or a graceful degradation message. Don't show raw error codes to end users.
  • Monitor OpenAI's status page actively. Set up an alert using status.openai.com's RSS feed or a service like UptimeRobot. You'll know the second they mark the incident resolved.
  • Don't spam the endpoints. Hammering the API during an outage doesn't speed up recovery. It makes the problem worse for everyone and might trigger rate limits on your key when service resumes.

What NOT to do: don't delete your API key and create a new one (it won't help). Don't open ten support tickets (see how to contact OpenAI support properly). Don't immediately upgrade your tier assuming it's a quota problem.

Questions people actually ask

Q: Is the OpenAI API down right now?

A: Check the live status monitor at the top of this page. It updates every 60 seconds. If it shows green for "API", the service is operational.

Q: How long do OpenAI API outages usually last?

A: Most last 15-45 minutes. Major incidents can run 2-3 hours. OpenAI's status page shows historical incident duration.

Q: Will OpenAI refund API credits for downtime?

A: OpenAI doesn't automatically refund for outages. You can request a review through platform.openai.com/account/billing if you lost significant credits during extended downtime.

Q: Should I build a fallback to another LLM provider?

A: For production-critical apps, yes. Services like Anthropic's Claude or Google's Gemini can serve as backup endpoints when OpenAI is down.

What to remember

  • The live status monitor on this site updates every 60 seconds from OpenAI's official feed
  • HTTP 500/503 errors point to OpenAI; 429 errors point to your rate limits
  • Test the `/v1/models` endpoint first to isolate the problem
  • Use exponential backoff and queues during outages — never spam retry
  • If basic troubleshooting fails, see ChatGPT not working — troubleshooting for broader connectivity issues

---

*openai-support.com is an independent resource, not affiliated with OpenAI Inc.*

Related help