How to raise your OpenAI rate limits

Where do you see your current limits?

In the OpenAI dashboard, under Settings, then Organization, then Limits. That page shows your tier and the limit for each model. OpenAI’s public table only covers a few model groups, so go by the dashboard.

How do you move up a tier?

Buy credits. Your tier goes up on its own once total credit purchases reach each threshold.

Tier	Total credit purchases	Monthly usage limit
Free	Allowed geography	$100
Build	$5	$500
Launch	$100	$5,000
Grow	$500	$200,000

Rates as listed on OpenAI’s Rate limits page, checked November 2, 2026.

From a standing start, Grow costs $500 in credits. The same Limits page has an Upgrade tier button in its Usage Tiers section. Credits are prepaid, and our billing guide covers what happens to them after you buy.

Per-model limits for every vendor are on our AI API rate limits page.

What if a tier still isn’t enough?

Grow is as far as self-serve goes. If your traffic keeps hitting ramp-rate limits, OpenAI points enterprise customers at Scale Tier, and at Reserved Tier for GPT-5.6 and later. Both are about capacity on eligible models, which means talking to sales.

Why are you getting 429s under your limit?

Ramping too fast is one cause. OpenAI returns a 429 with the code slow_down when traffic jumps, even while you’re inside your RPM and TPM. Its rule of thumb is that once you pass 1 million input tokens per minute, you add no more than 50% every 15 minutes.

Shared limits are another. Some model families draw on one pool, so traffic on one model can eat the budget of another. The Limits page marks which families share.

Long context has its own limit on models like GPT-5.5, separate from the regular one.

Then there are retries. Failed requests count toward your per-minute limit, so resending in a tight loop makes things worse.

Can you get more out of the limits you have?

Move anything that can wait to the Batch API. Batch has its own queue limit, based on input tokens queued per model, so it doesn’t compete with live traffic the same way.

Projects help too. Split work into projects and give each one a limit, and a runaway job can’t use up the whole organization’s budget.

When you do retry, follow the Retry-After header if it’s there and back off with a random delay if it isn’t.