Is OpenAI API down right now?
- No problems on the official status page, in our checks or in visitor reports.
Based on: OpenAI's official status page · our checks every minute · visitor reports. How we decide.
Is OpenAI API working for you?
In the last 15 minutes: 0 “down for me” and 0 “working for me” reports.
The OpenAI API is the developer service behind apps built on OpenAI's models. This page tracks the API components on OpenAI's status page (such as Chat Completions, Responses, Embeddings, Realtime and Batch) and checks that api.openai.com answers every minute.
Official status from OpenAI
OpenAI reports no open incidents for OpenAI API.
All 14 official components operational
| Chat Completions | Operational |
| Responses | Operational |
| Agents | Operational |
| Decisions | Operational |
| Fine-tuning | Operational |
| Embeddings | Operational |
| Images | Operational |
| Batch | Operational |
| Audio | Operational |
| Moderations | Operational |
| Realtime | Operational |
| Files | Operational |
| Login | Operational |
| Live | Operational |
Read . Open the official status page.
Our checks
Every minute we contact OpenAI API's public addresses from Cloudflare's network. Bot protection and missing API keys often cause 401, 403 or 404 replies; those still show the service is reachable.
| Address | Result | Time |
|---|---|---|
| API api.openai.com/v1/models | ✅ Reachable (401: needs a key, as expected) | 475 ms |
Last 24 hours
Not working for you? Try this
- Read the HTTP status code first401 means your API key is missing, wrong or revoked. 403 can mean the model or your country isn't available to your account. 429 means you're sending too much too fast, or your account is out of credit (the error message says which). 500 and 503 are problems on OpenAI's side and are worth retrying.
- Check billing and project limitsIf every request fails with a quota error, check billing and usage limits in the OpenAI platform dashboard. A spending limit on the project or organisation stops requests even while the API is healthy.
- Retry with exponential backoffDuring partial outages many requests still succeed. Retry failed requests after a short delay that doubles each time (1, 2, 4, 8 seconds), with a little randomness added, so your retries don't add to the load.
- Check only the endpoint you useOpenAI lists each API separately. If only Realtime or Batch is affected and you use Chat Completions or Responses, your app may be fine. The official components list on this page shows exactly which parts have problems.
- Rule out timeouts in your own codeLong answers take time to generate. If your HTTP client gives up after a few seconds, the request looks failed even though the API is working. Stream long responses or raise the timeout.
- Test with a minimal requestFrom the same server and with the same key, send the simplest request you can, such as listing models. If that works, the problem is in your request or code, not the API.
What to use while OpenAI API is down
- Anthropic API
Claude models. A common fallback for chat and coding workloads.No problems - Gemini
Google's models through the Gemini API.No problems - Mistral Le Chat
European provider with an API for its own models.No problems - DeepSeek
API for DeepSeek's chat and reasoning models.No problems
OpenAI API outages in the last 90 days
We haven't recorded any OpenAI API incidents in the last 90 days. We started tracking in October 2026.
Questions about OpenAI API outages
Is the OpenAI API down, or is it my code?
If our verdict is No problems and api.openai.com is answering, send a minimal request with your key from the same machine. If that works, the issue is in your code, key or account settings rather than the API.
What does a 429 error from OpenAI mean?
Either you've hit a rate limit (too many requests or tokens per minute) or your account has run out of credit or reached its spending limit. Rate limits clear by themselves within about a minute. Quota errors need a change in billing.
Does a ChatGPT outage mean the API is down?
Not always. ChatGPT and the API are listed separately by OpenAI and often have problems at different times. Check the components list on this page for the API parts you use.
Why does your check show a 401 error for api.openai.com?
We check the API without an API key, so it replies 'unauthorised' (401). That reply still proves the API is reachable. We never send paid requests.
OpenAI API is a product of OpenAI. IsAIDownNow is independent and not affiliated with OpenAI.