OpenAI API 429 error how to fix rate limits

You're hitting OpenAI's API and getting slammed with error 429. Your requests are failing, your application's broken, and you're stuck wondering why you can't use the tokens you're paying for.

OpenAI API 429 error how to fix rate limits

You're hitting OpenAI's API and getting slammed with error 429. Your requests are failing, your application's broken, and you're stuck wondering why you can't use the tokens you're paying for.

What's actually happening

Error 429 means you've exceeded OpenAI's rate limits — the maximum number of requests or tokens you can send in a given time window. OpenAI sets these limits per API key based on your usage tier, which ranges from free tier to tier 5. Every tier has different limits measured in requests per minute (RPM), tokens per minute (TPM), and tokens per day (TPD).

The actual error message looks like this: `Rate limit reached for requests` or `Rate limit reached for tokens`. You'll also see which limit you hit and when you can retry. For example, tier 1 users on GPT-4 get 500 RPM and 10,000 TPM. If you send 600 requests in 60 seconds, request 501 fails with 429. If you send massive prompts that consume 11,000 tokens in one minute, same error.

Your usage tier increases automatically as you spend more money with OpenAI — $5 spent gets you tier 1, $50 gets tier 2, $100 gets tier 3. You can check your current tier at platform.openai.com/settings/organization/limits. If you're seeing 429s constantly, you're either hitting your tier's ceiling or you need to implement better rate limiting in your code.

How to fix it

1. Check your current rate limits

Go to platform.openai.com/settings/organization/limits. You'll see your exact RPM, TPM and TPD limits for each model. Write these numbers down — you need them.

2. Implement exponential backoff

When you get a 429 error, the response includes a `retry-after` header telling you how many seconds to wait. Your code should catch 429 errors and retry with increasing delays. Start with 1 second, then 2, 4, 8, up to a maximum of 60 seconds. Most API libraries have built-in retry logic — enable it.

3. Add request batching

Instead of sending 1,000 individual requests per minute, batch them into fewer requests with multiple prompts. Use the batch API endpoint at platform.openai.com/batches if you're processing large volumes that don't need instant responses. Batch processing gives you 50% lower costs and separate rate limits.

4. Track your token usage

Count tokens before sending requests. Use OpenAI's tiktoken library to calculate exact token counts for your prompts and expected completions. If you're about to exceed your TPM limit, queue the request or wait. Don't just fire requests and hope.

5. Upgrade your usage tier

If you legitimately need more capacity, spend more money. Add $50 to your account to jump from tier 1 to tier 2, which typically doubles or triples your limits. Go to platform.openai.com/settings/organization/billing, add payment details, and increase your budget. Your tier updates within 24 hours of hitting the spending threshold.

If these steps don't work, check whether you have multiple API keys making requests. Each organization has combined limits across all keys — creating extra keys doesn't give you extra quota.

If that doesn't work

You might be seeing other API errors that look similar to 429 but have different causes. Error 401 means authentication failed — your API key is invalid. Error 500 or 503 means OpenAI's servers are temporarily down. Error 429 specifically with "insufficient_quota" means you've run out of credits entirely, which is different from hitting rate limits.

If you need higher limits immediately, contact OpenAI support at help.openai.com. Include your organization ID (found at platform.openai.com/settings/organization/general), your current tier, the specific limits you're hitting, and your use case. OpenAI rarely grants manual limit increases, but they'll review requests for legitimate high-volume applications. Expect 3-5 business days for a response.

Questions people actually ask

Q: Can I pay extra to remove rate limits completely?

A: No. Every tier has limits. The highest tier (tier 5, requiring $1,000+ in spending) has very high limits, but limits still exist. There's no unlimited tier.

Q: Why do I get 429 errors when my dashboard shows I haven't used many tokens?

A: The dashboard shows total usage. Rate limits are per-minute or per-day windows. You might be bunching all your requests into short bursts that exceed the per-minute cap even though your daily total is low.

Q: Does creating multiple API keys give me more quota?

A: No. Rate limits apply to your entire organization, not individual keys. Ten keys share the same pool of requests and tokens.

Q: How long until my rate limit resets?

A: RPM limits reset every minute. TPM limits reset every minute. TPD limits reset every 24 hours. If you hit a limit at 2:34 PM, you can try again at 2:35 PM for per-minute limits.

What to remember

  • Check your exact rate limits at platform.openai.com/settings/organization/limits
  • Implement exponential backoff with retry-after headers in your code
  • Use tiktoken to count tokens before sending requests
  • Batch processing gives you separate limits and 50% cost savings
  • Spending more money automatically increases your tier and limits within 24 hours

---

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

Related help