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

You send a request to the OpenAI API and get a 503 error, a timeout, or nothing at all. Before you rewrite your retry logic or file an urgent ticket, check if it's actually OpenAI having problems — you can verify in unde

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

You send a request to the OpenAI API and get a 503 error, a timeout, or nothing at all. Before you rewrite your retry logic or file an urgent ticket, check if it's actually OpenAI having problems — you can verify in under 30 seconds. This guide shows you exactly how to confirm an outage, isolate the issue, and keep your application running when the API goes sideways.

How to check if OpenAI is actually down

Start with the live OpenAI status monitor on this site — it updates every 60 seconds from OpenAI's official status feed and shows component-level health for the API, ChatGPT, and other services. Look specifically at the "API" row. Green means operational, yellow indicates degraded performance, red means a full outage.

Cross-reference with status.openai.com directly. OpenAI posts incident reports there with timestamps, affected endpoints, and updates as they work through problems. Check the "API" section specifically — sometimes ChatGPT works fine while `/v1/chat/completions` returns errors.

Watch for patterns in your error codes. A genuine API outage typically shows 503 Service Unavailable, 529 Overloaded, or connection timeouts across all your requests. If you're seeing 429 Rate Limit Exceeded or 401 Unauthorized, that's not an outage — those are account-specific issues covered in OpenAI API error 429 and other API errors.

Third-party monitors like Downdetector spike when enough people report problems simultaneously. Combine that with the live OpenAI status here and the official status page. If all three sources show issues at the same time, it's OpenAI's infrastructure, not yours.

Is it OpenAI or is it you

  • Test the endpoint directly. Run a simple curl command against `https://api.openai.com/v1/models` with your API key. If this succeeds but your application fails, the problem lives in your code or infrastructure.
  • Check your API key permissions. Log into platform.openai.com, navigate to Settings → API keys, and verify your key hasn't been revoked or rate-limited. Create a new key and test with that — sometimes keys get silently restricted.
  • Review your network path. Try the same request from a different network — your mobile hotspot, AWS, or a VPS. If it works elsewhere, your office network, VPN, or hosting provider might be blocking OpenAI's IP ranges.
  • Isolate to specific models. Test `gpt-3.5-turbo` and `gpt-4` separately. During partial outages, newer models sometimes fail while older ones stay up. Switch models temporarily if one works.
  • Examine request patterns. Pull your logs for the last 100 requests. If 95 succeed and 5 fail with 503 errors, that's API instability, not a full outage. If everything fails, check steps 1-3 again.
  • Verify account billing. Go to platform.openai.com → Settings → Billing. An unpaid invoice or expired card triggers hard stops that look like API errors. Add credits or update payment info.

What to do during an outage

  • Implement exponential backoff immediately. When you hit errors, wait 1 second, then 2, then 4, then 8, up to 60 seconds between retries. This code pattern should already exist in your application — if it doesn't, add it now before the next outage.
  • Queue failed requests. Don't drop them. Store them in Redis, PostgreSQL, or a message queue with timestamps. Process the queue when the API recovers. Set a maximum queue depth — 10,000 requests is reasonable, 100,000 will kill your database.
  • Switch to cached responses. If you're running chatbots or content generation, serve cached responses for common queries during outages. Better than showing users error messages for repetitive questions.
  • Monitor your retry budget. Count total retries per hour. If you're sending 1,000 requests and 900 fail, stop retrying and wait 15 minutes. Hammering a down API makes recovery slower for everyone.
  • DO NOT delete your account, create new API keys in panic, or spam the support form. These actions don't fix outages. They create more work for you when the API recovers. If you need to reach OpenAI support for genuine account issues, see How to contact OpenAI support.

Questions people actually ask

Q: Is the OpenAI API down right now?

A: Check the live OpenAI status monitor at the top of this site — it shows real-time API health updated every 60 seconds. Green means working, red means down.

Q: How long do OpenAI API outages usually last?

A: Most last 10-45 minutes. Major incidents can run 2-4 hours. Status.openai.com posts estimated recovery times during active incidents.

Q: Will OpenAI refund API credits for downtime?

A: OpenAI's terms don't guarantee refunds for outages. You can request credit through platform.openai.com → Help → Messages, but approval isn't automatic.

Q: Can I switch to Azure OpenAI during outages?

A: Azure OpenAI runs on separate infrastructure, so it sometimes stays up when api.openai.com goes down. You need an Azure account and separate API keys — setup takes hours, not minutes.

What to remember

  • Verify outages with three sources: this site's live status, status.openai.com, and your own error logs
  • Exponential backoff and request queueing prevent cascading failures in your application
  • Test specific models separately — partial outages affect different endpoints differently
  • Account billing issues and rate limits aren't outages, they're configuration problems
  • During a real outage, waiting is faster than retrying aggressively

---

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

Related help