Why does OpenAI API keep returning 429 errors

You're calling the OpenAI API and getting slammed with 429 errors over and over. Your code worked fine yesterday, or you just upgraded your tier, but the requests still fail with "Rate limit reached" or "You exceeded you

Why does OpenAI API keep returning 429 errors

You're calling the OpenAI API and getting slammed with 429 errors over and over. Your code worked fine yesterday, or you just upgraded your tier, but the requests still fail with "Rate limit reached" or "You exceeded your current quota."

What's actually happening

The 429 error means OpenAI's servers are refusing your request because you've hit a limit — but there are three completely different types of limits that all return the same error code.

Rate limits control how many requests you can make per minute (RPM) and how many tokens per minute (TPM). Even paid accounts have caps: GPT-4 on tier 1 maxes out at 500 RPM and 10,000 TPM. If you're firing off requests in a loop without delays, you'll blow through these limits in seconds. The error message usually says "Rate limit reached for requests" or includes your specific RPM/TPM numbers.

Usage quotas are different — they're monthly spending caps OpenAI sets on your account. New accounts often start with $100/month limits. If you've burned through your monthly allowance, every request returns 429 with "insufficient_quota" in the error details. This happens even if you're only making one request per hour.

Organisation-level limits kick in when multiple team members share an API key. The limits apply to the entire organisation, not individual users. One person running a heavy workload can block everyone else's requests.

How to fix it

1. Check which limit you actually hit

Log into platform.openai.com/account/limits. This page shows your current tier, RPM/TPM limits for each model, and how close you are to hitting them. If you see "100% of limit used" next to requests or tokens, that's your rate limit problem.

Then check platform.openai.com/usage. Look at your spending this month versus your quota. If you're at $98 out of $100, you've hit your usage quota, not your rate limit.

2. Add exponential backoff to your code

Most 429 errors from rate limits happen because you're sending requests too fast. Add a retry loop that waits longer after each failure:

```python

import time

for attempt in range(5):

try:

response = client.chat.completions.create(...)

break

except openai.RateLimitError:

wait_time = (2 ** attempt) + random.random()

time.sleep(wait_time)

```

This waits 1 second, then 2, then 4, then 8 seconds between retries. The random addition prevents multiple scripts from retrying in lockstep.

3. Request a tier increase

If you're consistently hitting rate limits and you've been using the API for at least 7 days with successful payments, go to platform.openai.com/settings/organization/limits and click "Request tier increase." OpenAI reviews these manually — expect 2-5 business days. They look at your payment history and usage patterns. Accounts with failed payments or brand-new accounts usually get rejected.

4. Increase your usage quota

For quota issues, go to platform.openai.com/settings/organization/billing/limits and adjust your monthly budget. You can set this to any amount, but OpenAI might require additional payment verification for limits above $1,000 if your account is new.

If that doesn't work

You might be hitting the 429 error even after fixing rate limits because of token-per-day (TPD) limits on specific models. GPT-4 Turbo has separate daily token caps that don't show up in the main limits dashboard. Check the API response headers — look for `x-ratelimit-remaining-tokens` and `x-ratelimit-reset-tokens`.

If your organisation has multiple developers, audit who's using the shared API key. One person running an infinite loop can lock out everyone else. Create separate projects with isolated API keys at platform.openai.com/settings/organization/projects.

When you've verified your limits are sufficient but 429s persist, contact OpenAI support through help.openai.com. Include your organisation ID, the exact timestamp of failed requests, and the full error response body. Response times average 1-3 business days for API technical issues.

Questions people actually ask

Q: I just added $50 to my account but I'm still getting 429 errors — why?

A: Added credits don't automatically increase your usage quota. Go to billing settings and manually raise your monthly limit to match your new balance.

Q: Can I pay extra to skip rate limits entirely?

A: No. Even tier 5 accounts (requiring $1,000+ monthly spend for 6+ months) have rate limits — they're just 100x higher than tier 1. You can't buy your way to unlimited requests.

Q: The error says "insufficient_quota" but my billing page shows $200 available — what's wrong?

A: Check if you have unexpected charges pending. OpenAI sometimes places holds on accounts with disputed payments or unusual usage spikes. Pending charges count against your quota before they clear.

What to remember

  • Check platform.openai.com/account/limits and /usage to identify whether you're hitting rate limits or quota limits
  • Add exponential backoff with random jitter to all API calls — never retry immediately
  • Usage quotas and rate limit tiers are separate settings that both need adjustment as you scale
  • New accounts need 7+ days of payment history before tier increases get approved
  • Organisation-wide API keys share limits across all users — create separate project keys for isolation

---

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

Related help