ChatGPT errors FAQ: every question answered

Most ChatGPT errors come from one of three things: an OpenAI-side problem (outage, high traffic, "at capacity", 500/503), a rate limit because you sent too many messages (429, "too many requests"), or something on your side (network, VPN, browser cache, extensions, or a blocked IP causing error 1020). First check the OpenAI status page; if it's green, the fix is usually on your side.

It depends on the error. Server errors like "at capacity", 500, 502 and 503 are on OpenAI's side. Rate limits (429, "too many requests") are triggered by how fast you send messages. Access errors such as 1020, plus most "something went wrong" and login issues when status is green, are usually on your side — network, VPN, cache or extensions.

Check the official status page at status.openai.com, which lists incidents, degraded performance and maintenance. As a cross-check you can look at third-party outage trackers like Downdetector. If both are clear, the problem is more likely local — try a refresh, incognito, a different network, or turning off your VPN.

It's a generic error that can be caused by a server hiccup, a long or stuck conversation, browser cache or cookies, an extension, or a network issue. Refresh, start a new chat, try incognito, clear cache, or switch network. If status.openai.com shows an incident, just wait it out.

A network error often means the connection dropped mid-reply or a request couldn't reach OpenAI. Check your internet, refresh and retry, turn off your VPN or proxy, switch network, disable extensions, and shorten very long prompts that may time out. If the status page reports an incident, the network error may actually be on OpenAI's side.

429 means you've sent too many messages in a short time and hit a rate limit. Wait a few minutes, slow down, close extra ChatGPT tabs or devices using the same account, and avoid rapid retries. On the API, reduce request frequency and use exponential backoff.

Error 1020 is a Cloudflare access-denied block tied to your IP or network. Turn off your VPN or proxy (or switch servers), clear ChatGPT cookies, try another network or device, update your browser, and make sure your date and time are correct. It's almost always a your-side or network issue, not your account.

These are OpenAI-side server errors: 500 is an internal server error, while 502, 503 and "at capacity" mean the servers are temporarily overloaded or unavailable. You can't fix them from your end — wait, refresh after a short pause, and check the status page. Retrying aggressively can make it worse.

"At capacity" means OpenAI's servers are handling more demand than they can serve at that moment, so new requests are temporarily turned away. It's an OpenAI-side condition, not something on your device. Wait a little and try again, check status.openai.com, and keep working in another model meanwhile.

For login issues, clear cookies and cache, try incognito, disable extensions, reset your password, confirm your email is verified, and check that status.openai.com isn't reporting an authentication incident. On mobile, update or reinstall the app and sign in again. Only ever enter your password on the official ChatGPT login page.

App errors are usually fixed by updating the app, checking your connection, signing out and back in, clearing the app cache or reinstalling, and confirming the OpenAI status page is green. If only the app fails, try the web app at chatgpt.com to isolate whether it's the app, your account or OpenAI.

Usually no. Most errors are temporary server issues, rate limits, or local network and browser problems. A 1020 block is about your IP or network, not a ban. If you genuinely can't access your account after trying the standard fixes, contact OpenAI support.

Sometimes. A single refresh or a new chat clears many transient "something went wrong" errors. But for server errors (500/503/at capacity) or rate limits (429), rapid retrying doesn't help and can extend the limit — wait a bit, then try again.

A reply that cuts off mid-stream is often a network drop, a timeout on a very long generation, or a brief server issue. Try asking it to "continue", refresh and resend, shorten the prompt, switch network, or disable a VPN. If the status page reports an incident, it may be an OpenAI-side problem rather than your connection.

Often, yes — for your-side errors. Stale cookies or a corrupted cache can cause "something went wrong", login loops and access errors. Trying an incognito window first tells you whether cache or extensions are the cause; if it works there, clear cache and cookies or disable extensions in your normal browser.

Yes. A VPN, proxy or shared IP can trigger access blocks such as error 1020, slow or unstable connections, and login problems. Turning the VPN off or switching to a different server or network often resolves these. If turning it off fixes the issue, the block was tied to that IP or network.

A rate limit (429, "too many requests") affects only you because you sent requests too quickly, and it clears after you slow down and wait. An outage affects many users at once, shows up on status.openai.com, and is on OpenAI's side — there's nothing to fix locally, so you wait for them to recover.

1) Check status.openai.com. 2) If green, refresh and try a new chat. 3) Open an incognito window (rules out cache and extensions). 4) Switch network or disable VPN (rules out 1020 and blocks). 5) If you hit a rate limit, wait. 6) Meanwhile, keep working in another model via MultipleChat.

They can. The API has its own rate limits (429) and can return server errors (500/502/503) during incidents, which the status page also covers. For the API, the standard mitigations are to slow down, use exponential backoff on retries, and check the status page rather than assuming a code bug on your side.

Use a multi-model workspace like MultipleChat: when ChatGPT is erroring or at capacity, you can switch to Claude, Gemini or another model and keep going for free, instead of waiting for OpenAI to recover. It also lets you compare answers across models.