A 429 from the OpenAI API means you've hit a limit — but which one determines the fix, and retrying the wrong way can waste time or even run up cost for nothing. Here's how to tell them apart.
First, check which kind of 429 you're actually looking at
OpenAI's own troubleshooting documentation groups 429s into three causes: rate limiting (too many requests or tokens in a given period), billing issues (exhausted prepaid credit or a hit spend limit), and usage quota (an OpenAI-assigned usage limit or an enforced organization/project spend ceiling). The error body and message text tell you which one you hit — read it before choosing a fix, because a rate limit and a billing block need completely different responses.
If it's a rate limit: back off, don't just retry
Space requests out over time instead of firing bursts, add exponential backoff with jitter to your retry logic, and honor the Retry-After response header when the API returns one. Trimming unnecessary content from prompts and setting a sensible output-token cap also reduce how fast you burn through your per-minute allowance.
If it's a billing or quota error: retrying won't help
OpenAI is explicit that retrying a billing, spending, or quota error does not restore access on its own. Add credit or raise your spend limit in your billing settings, request a higher approved usage limit if you've hit an OpenAI-assigned ceiling, or wait for a monthly reset if one applies to your account.
Growing out of your current limits for good
Rate limits rise automatically as your account moves through OpenAI's usage tiers with cumulative spend — see our guide on why an OpenAI bill increases for how that progression works. If you've outgrown your current tier's limits even at normal spend, that's the lever to look at before assuming something is broken.