Check the status page
Open status.openai.com. An incident or degraded performance confirms it's OpenAI's side, not yours.
ChatGPT errors, explained
"At capacity" — and the 503 "service unavailable" status code behind it — means OpenAI's servers are temporarily overloaded or under maintenance. It's an OpenAI-side problem, not your account, network or device, so there's nothing to fix locally. This independent guide explains what it means, why hammering retry makes things worse, and how to keep working for free while the servers recover.
What it means
These are all OpenAI-side server errors. You can't fix them locally — the table shows what each means and what to do.
| What you see | Meaning | Whose side | What to do |
|---|---|---|---|
| "At capacity" | Servers temporarily overloaded and can't take more requests | OpenAI | Wait, refresh after a pause, check the status page, work elsewhere. |
| 503 | Service unavailable — overloaded or under maintenance | OpenAI | Same as above. Don't hammer retry; let the servers recover. |
| 500 | Internal server error on OpenAI's side | OpenAI | Wait and retry after a pause; check status.openai.com. |
| 502 | Bad gateway — an upstream server error on OpenAI's side | OpenAI | Wait and retry after a pause. See the error codes guide. |
Don't just sit there
A 503 / "at capacity" is OpenAI's servers, not you — so there's no local fix and waiting is the only option on ChatGPT. But you don't have to stop working. A multi-model workspace lets you switch to another AI and finish what you were doing, then go back to ChatGPT once it recovers.
When ChatGPT is at capacity or returning 503, switch to Claude, Gemini or another model in the same place and keep going for free — no waiting for OpenAI's servers to recover. Compare answers across models too.
OfficialThe first place to check: live incidents, outages and maintenance.
OfficialOpenAI's own troubleshooting and error articles.
GuideHow to confirm an outage and what to do.
What to do
You can't fix a server-side error, but you can handle it well. Work through these so you don't make it worse — and so you keep moving.
Open status.openai.com. An incident or degraded performance confirms it's OpenAI's side, not yours.
Give it a pause and refresh after a short wait. A 503 is temporary and usually clears once load drops.
Aggressive retrying adds load to already-overloaded servers and can make the problem worse. One attempt after a pause is enough.
Clearing cache or switching network won't help a server-side 503 — those fixes are for your-side errors like 1020.
Keep working in MultipleChat while ChatGPT recovers, instead of waiting and watching the screen.
If it's a bigger incident, check the status page for updates and return once OpenAI marks it resolved.
Related guides: all error codes · is ChatGPT down? · too many requests · something went wrong
All error guides
FAQ
Short answers on what "at capacity" and 503 mean, why retrying hurts, and how to keep working. Verify current behaviour on OpenAI's official pages.
"At capacity" means OpenAI's servers are temporarily overloaded and can't take on more requests right now. It's an OpenAI-side issue, not a problem with your account or device. The usual remedy is to wait a little and try again later.
A 503 is the "service unavailable" HTTP status code. For ChatGPT it generally means the servers are temporarily overloaded or under maintenance — the same situation often shown as "at capacity". It's an OpenAI-side error you can't fix locally.
They overlap. "At capacity" is a friendly message and 503 is the underlying "service unavailable" status code; both point to OpenAI's servers being overloaded or unavailable. In practice you treat them the same way: wait, don't hammer retry, and check the status page.
You can't fix a 503 from your end because it's a server-side problem. Wait a bit, refresh after a pause rather than hammering retry, and check status.openai.com for an incident. Meanwhile, switch to another model so you can keep working.
When servers are already overloaded, aggressive retrying adds more load and can make the situation worse, not better. Give it a pause between attempts instead of hammering refresh, and let OpenAI's servers recover.
There's no fixed duration — it depends on the load or maintenance on OpenAI's side and isn't published. It's usually temporary. Check status.openai.com for updates on any incident, wait, and try again after a pause rather than retrying constantly.
No. A 503 / "at capacity" is an OpenAI-side server issue, not something wrong with your account, network or device. That's why local fixes like clearing cache or switching network won't resolve it — the only real options are to wait and to work elsewhere.
Check status.openai.com. If OpenAI reports an incident or degraded performance, that confirms it's their side and explains the "at capacity" or 503 message. If status is green and you still see errors, the cause may be something more local — see the error codes guide.
Use a multi-model workspace like MultipleChat: while ChatGPT is at capacity or returning 503, switch to Claude, Gemini or another model and keep going for free instead of waiting for OpenAI's servers to recover. It also lets you compare answers across models.