OpenAI API 429 rate limit how to fix it

You're calling the OpenAI API and getting slammed with error 429 — "Rate limit reached" — even though you're barely making requests. This hits developers mid-project, blocks production apps, and makes no sense when you'r

OpenAI API 429 rate limit how to fix it

You're calling the OpenAI API and getting slammed with error 429 — "Rate limit reached" — even though you're barely making requests. This hits developers mid-project, blocks production apps, and makes no sense when you're under your token quota.

What's actually happening

Error 429 means you've exceeded OpenAI's rate limits, but there are actually three different limits working against you: requests per minute (RPM), tokens per minute (TPM), and requests per day (RPD). You can max out any one of these and get rate-limited even if the others have headroom.

Free tier accounts get brutally low limits — 3 RPM and 40,000 TPM on GPT-3.5-turbo, 200 TPM on GPT-4. If you're on a paid tier, your limits scale with how much you've spent historically: $5 total spend gets you tier 1 with 500 RPM on GPT-4, $50 gets tier 2 with 5,000 RPM. The system doesn't care about your current balance — it looks at cumulative API spending since your account opened.

The error response looks like this:

```

{

"error": {

"message": "Rate limit reached for gpt-4 in organization org-xxx on requests per min. Limit: 200 / min.",

"type": "rate_limit_error",

"code": "rate_limit_exceeded"

}

}

```

That message tells you exactly which limit you hit and what your ceiling is.

How to fix it

1. Check your current limits

Go to platform.openai.com/account/limits. This page shows your exact RPM, TPM and RPD caps for every model. If you see "3 requests per minute" on GPT-4, you're still on free tier limits despite having credits.

2. Upgrade your usage tier

Add $5-10 to your account at platform.openai.com/account/billing/overview, then actually spend it on API calls. Make real requests until you cross the tier threshold — depositing money doesn't count, only spent tokens do. Tier 1 needs $5 spent total, tier 2 needs $50, tier 3 needs $100. Check the limits page again after 10-15 minutes.

3. Implement exponential backoff

Your code needs retry logic. When you hit 429, wait 1 second and retry. If it fails again, wait 2 seconds, then 4, then 8. OpenAI's Python library does this automatically if you wrap calls in a try/except block, but you'll need manual implementation in other languages. The error response includes a `Retry-After` header — use that wait time if present.

4. Batch your requests smarter

If you're sending 100 requests instantly, you'll hit RPM limits even with high TPM capacity. Spread requests across 60 seconds using a queue. Tools like `asyncio` in Python or promise batching in Node.js let you control request timing without blocking your app.

5. Switch models strategically

GPT-3.5-turbo has 10x higher rate limits than GPT-4 at every tier. If your task works on 3.5, route those requests there and save GPT-4 calls for complex queries. You can check different model limits at platform.openai.com/docs/guides/rate-limits/usage-tiers.

If rate limits stay low after 24 hours of tier-qualifying spend, your account might need manual review. Submit a request at help.openai.com/en/articles/6891753-how-do-rate-limits-work explaining your usage tier and historical spend.

If that doesn't work

You need to contact OpenAI support with specific details. Go to help.openai.com and click "Messages" in the bottom right, or email them through the platform. Include your organization ID (starts with `org-`, found at platform.openai.com/account/organization), the exact error message with timestamps, your current usage tier from the limits page, and your total historical spend. They typically respond in 2-4 business days for API issues.

For urgent production needs, some developers spin up multiple organizations under different emails — each gets separate rate limits — but OpenAI's terms discourage this and they can link accounts. The legitimate path is contacting OpenAI support for a tier review if you have genuine high-volume needs.

Questions people actually ask

Q: I added $100 to my account but still get 429 errors — why?

A: Credits don't count toward usage tiers. You have to actually spend those tokens on API calls. The tier system looks at historical consumption, not your balance.

Q: Can I pay to instantly increase my rate limits?

A: No direct payment option exists. You advance tiers by spending cumulatively over time — $5 spent total for tier 1, $50 for tier 2, $100 for tier 3. There's no skip-ahead purchase.

Q: How long does it take for tier upgrades to apply?

A: Usually 10-15 minutes after you cross the spending threshold. Check platform.openai.com/account/limits to confirm. If it's been 24 hours, contact support.

Q: Do rate limits reset daily?

A: RPM and TPM reset every minute on a rolling window. RPD limits reset at midnight UTC. If you hit daily limits, you're completely blocked until the next day.

What to remember

  • Check platform.openai.com/account/limits for your exact caps before debugging
  • You must actually spend money on API calls, not just deposit it, to upgrade tiers
  • Implement exponential backoff with 1-2-4-8 second delays in your retry logic
  • Free tier gets 3 RPM on GPT-4 — completely unusable for real apps
  • Different models have different limits even at the same tier — use GPT-3.5 when possible

Related help

---

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

Related help