ChatGPT “Error 429: Too Many Requests” Fix (Web & API) 2026

A 429 from ChatGPT is a rate limit. It means too many requests reached the service in a short window, either from you, your network, or in the case of developers, your API key. The fix depends on which of those is sending the traffic.

Error snapshot
Exact error code	429
Error message	Too Many Requests. You are sending requests too quickly. (HTTP 429)
Applies to	ChatGPT web app, mobile app, and the OpenAI API
Environment	Browser, app, and developer integrations
Severity level	Low temporary throttle, clears on its own or after a quota top up
Quickest fix	Wait, refresh, and send fewer requests per minute. API users check quota and billing.

What the error actually means

HTTP 429 is a standard too many requests response. ChatGPT returns it when the volume of requests crosses a limit. On the consumer app this can happen during very heavy usage, rapid repeated sends, or when many people share one office or campus network and one public address.

For developers calling the API, 429 has a second meaning. It can signal that you have exceeded the requests per minute or tokens per minute limit for your tier, or that your account has run out of paid quota. The response headers usually tell you which.

Before you start

Step by step fixes

The first three steps fix the web app. The rest are for API and integration users.

#	Step
1	Wait, then refresh. Give it thirty to sixty seconds and reload the page or app. A short pause is often all a consumer 429 needs.
2	Slow your sending. Stop firing messages back to back. Let one response finish before you send the next, and avoid spamming the regenerate button.
3	Switch network or sign in fresh. If you are on shared office or campus Wi-Fi, try mobile data, since the limit can be tied to the shared public address. Signing out and back in also clears a stuck session.
4	API users, check usage and billing. Open your usage dashboard. If you are out of quota, the 429 is a billing limit, not a speed limit. Add credit or raise your limit.
5	Read the rate limit headers. Inspect the response for the retry and rate limit headers. They tell you exactly how long to wait and which limit you crossed.
6	Add exponential backoff. In code, catch the 429, wait, then retry with a delay that doubles each time up to a cap. This is the standard, supported way to handle rate limits.
7	Batch and reduce calls. Combine small requests, cache repeated answers, and remove unnecessary calls so your steady request rate sits under the limit.

The technical context (the why)

Rate limiting protects a shared service from being overwhelmed and keeps response times fair for everyone. Without it, a handful of heavy users or a runaway script could degrade the experience for millions. The 429 is the polite way the service asks a noisy client to wait.

For API integrations, the limits also map to your plan. Each tier allows a certain number of requests and tokens per minute, and a certain amount of paid usage. When you cross either, the server answers with 429 and, in the headers, tells your code how to recover. Handling that response with backoff is part of building a reliable integration.

Key points to remember

• On the web app, waiting and slowing down clears most 429s.
• On the API, a 429 can mean rate limit or exhausted quota, so check both.
• Never retry instantly in a loop, use a growing backoff delay.
• Shared networks can hit the limit collectively, so switching networks can help.

Still broken? Your safety net

If you are a single user on a normal connection and 429 persists for a long time, check the OpenAI status page for an incident, then sign out, clear the site data, and sign back in. For API accounts where billing is current and limits are not exceeded, contact OpenAI support with a sample request id from the failed call.

OpenAI help centerMore Error Fix guides