zumatools.

Popular right now

Add Prefix / Suffix to LinesPaste one item per line, then add text around each line, join them with a separator,…Tally CounterTap the big card to count up, use the minus button to correct mistakes, and add as…Love CalculatorEnter two names and get a playful compatibility score — the same pair of names always…Photo to Pencil SketchConvert any picture into a hand-drawn pencil sketch directly in your browser, with…Tic Tac ToePlay against a friend on the same device or challenge the computer on easy or…Random Pair GeneratorPaste names below and split them into random pairs or groups of three, shuffled fairly…
Free tools. Right here in your browser.esc to close

OpenAI · Developer API

Is OpenAI API down right now?

The OpenAI API is OpenAI’s developer platform for building with its models. It includes endpoints such as Responses, Chat Completions, Embeddings, Images, Audio, Realtime, Batch and Files.

Live status · OpenAI API

Checking live status…

Reading the official status page, our reachability checks and recent user reports.

Official status page
—status.openai.com
Our reachability check
—Checked every minute from our server
User reports
—From visitors to this page

Having trouble with OpenAI API?

Tell us what’s wrong. One report per person every 30 minutes; nothing else is collected.

OpenAI API in the last 24 hours

Loading the last 24 hours…

Reachability (our checks)No data
  • Reachable
  • Partly reachable
  • Not reachable
  • No data
Response timeNo data
User reportsNo data

If OpenAI API isn’t working for you

When the live checks above look normal, the problem is usually on your side or limited to one feature. These are the fixes that help most often with OpenAI API:

  1. Check the specific API component

    The status page lists API components separately: Responses, Chat Completions, Embeddings, Images, Audio, Realtime, Batch, Fine-tuning, Files and more. Match the failing endpoint to its component — an incident on one endpoint doesn’t mean the whole API is down.

  2. 429: rate limit vs. quota

    A 429 can mean several things. “Rate limit reached for requests” means you’re sending requests too fast, so follow the Retry-After header and ramp up gradually. Quota, credit-balance and spend-limit errors are billing limits, and retrying won’t fix them until you add credits or raise the limit.

  3. 500 and 503 errors

    A 500 (“The server had an error while processing your request”) is a problem on OpenAI’s side — retry after a short wait. A 503 means the model or service is short on capacity; follow Retry-After when it’s present, then retry.

  4. Retry with exponential backoff

    For 429 rate-limit errors and 5xx errors, retry with exponential backoff and jitter instead of resending immediately. Hammering the API with the same request can make the problem last longer.

  5. 401 and 403 are not outages

    A 401 means authentication failed: a wrong or revoked API key, the wrong organization, or an IP address blocked by your allowlist. A 403 “Country, region, or territory not supported” means the API isn’t available where the request comes from.

Official OpenAI API sources

Questions about OpenAI API outages

Is the OpenAI API down or am I rate limited?

Read the 429 message. Rate-limit messages mean slow down and retry; quota or spend-limit messages mean a billing limit. Only 5xx errors across many requests, together with an incident on the status page, point to an outage. This page combines the official status page, our own reachability checks every minute and user reports.

Does status.openai.com cover the API and ChatGPT?

Yes. It has separate groups for APIs, ChatGPT, Codex and others. Check the APIs group for your endpoint, because a ChatGPT incident doesn’t always affect the API.

What does a 503 from the OpenAI API mean?

It means the model or service you called is short on capacity at the moment. Wait for the Retry-After period if one is given, then retry with backoff. It is often temporary and limited to one model.

Should I retry every failed request?

No. Retry 429 rate-limit errors and 5xx errors with exponential backoff. Don’t blindly retry 401 or 403 errors, or quota and spend-limit errors, because they won’t succeed until you fix the key, the region or the billing.

How this page decides

The answer at the top combines three signals: what OpenAI reports on its official status page, our own check that OpenAI API is reachable (every minute, with response time), and reports from people on this page. Problems are flagged when the official source reports an incident, our checks start failing, or reports rise well above the usual level for OpenAI API. It’s an independent page — not affiliated with OpenAI. See all AI services.

Check another AI service