OpenAI API error 429 rate limit exceeded

You're making API calls and suddenly hit a wall: Error 429: Rate limit exceeded. Your application stops working, requests fail, and you're stuck wondering if you're being throttled or if you've maxed out your quota.

OpenAI API error 429 rate limit exceeded

You're making API calls and suddenly hit a wall: `Error 429: Rate limit exceeded`. Your application stops working, requests fail, and you're stuck wondering if you're being throttled or if you've maxed out your quota.

What's actually happening

Error 429 means OpenAI's servers are deliberately rejecting your requests because you've exceeded one of several rate limits. This isn't a bug — it's OpenAI enforcing usage caps to prevent system abuse and ensure fair access across all users.

OpenAI sets three types of limits on every account: requests per minute (RPM), tokens per minute (TPM), and tokens per day (TPD). Your limit depends on your usage tier, which increases automatically as you spend more money over time. A brand-new account might get 500 RPM and 200,000 TPM on GPT-4, while established accounts with higher spending get 10,000 RPM and 5,000,000 TPM. If you hit any of these limits, you get a 429 error with a message like `Rate limit reached for requests` or `Rate limit reached for tokens`.

The confusing part: you might have plenty of quota (prepaid credits) remaining but still hit rate limits. These are separate systems. Rate limits control *how fast* you can make requests. Quota controls *how much* you can spend. You need both to keep your API running.

How to fix it

1. Check your current limits

Log into platform.openai.com, click Settings in the left sidebar, then Limits. You'll see your exact RPM, TPM, and TPD limits for each model. Compare these to what your application is actually requesting.

2. Implement exponential backoff

When you hit a 429 error, wait before retrying. Start with a 1-second delay, then double it with each failure (2 seconds, 4 seconds, 8 seconds). Most OpenAI client libraries have built-in retry logic — make sure it's enabled. Here's what matters: the `Retry-After` header in the 429 response tells you exactly how long to wait. Respect it.

3. Spread out your requests

If you're sending 100 requests at once, you'll slam into your RPM limit immediately. Add delays between batches — if your limit is 500 RPM, send no more than 8 requests per second. Use a queue system to smooth out traffic spikes instead of firing everything simultaneously.

4. Request a limit increase

On the Limits page, click Request increase next to the model you need. Fill out the form explaining your use case, expected traffic, and why current limits aren't sufficient. OpenAI typically responds within 3-5 business days. Be specific — "We're processing customer support tickets and need 2,000 RPM during business hours" works better than "We need more access."

5. Upgrade your tier automatically

Your tier increases as you spend more. If you're on Tier 1 (under $5 spent), reaching Tier 2 ($50+ spent) doubles most limits. Check platform.openai.com/settings/organization/limits to see what you need to spend to reach the next tier. There's no manual upgrade — it happens automatically once you hit the spending threshold.

If that doesn't work

You've implemented backoff, spread requests, and still hit 429s constantly. Here's what to check: make sure you're not counting tokens incorrectly. A conversation with 20 messages might consume 5,000 tokens — count input *and* output tokens together. Use OpenAI's tokenizer tool at platform.openai.com/tokenizer to verify.

If you need immediate relief and can't wait for a limit increase, consider these alternatives: batch multiple operations into single requests where possible, use a smaller model (GPT-3.5-turbo has higher limits), or implement caching to avoid re-processing identical requests.

For persistent issues after trying everything above, contact OpenAI support through the chat widget on platform.openai.com. Include your organization ID, the specific model causing 429s, your current limits from the Limits page, and what you've already tried. Don't expect instant responses — OpenAI support typically takes 24-72 hours for technical issues. You can find more details about other API errors that might look similar but have different causes.

Questions people actually ask

Q: Does error 429 mean I'm out of credits?

A: No. Error 429 is about rate limits (speed), not quota (money). Check platform.openai.com/settings/organization/billing to see your actual credit balance. You can have $100 in credits and still hit 429 if you're making requests too fast.

Q: How long do I have to wait after getting a 429?

A: Check the `Retry-After` header in the error response — it tells you exactly how many seconds to wait. If there's no header, start with 1 second and use exponential backoff. Rate limits reset every minute for RPM/TPM and daily for TPD.

Q: Will paying for ChatGPT Plus increase my API limits?

A: No. ChatGPT Plus ($20/month) and API access are completely separate. API limits depend only on your API spending tier and approved increase requests.

What to remember

  • Error 429 means you're making requests too fast or using too many tokens too quickly — not that you're out of money
  • Check platform.openai.com/settings/organization/limits to see your exact RPM, TPM, and TPD caps
  • Implement exponential backoff with the `Retry-After` header value
  • Request limit increases through the Limits page with specific justification
  • Your tier (and limits) increases automatically as you spend more on API usage

---

*openai-support.com is an independent resource, not affiliated with OpenAI Inc.*

Related help