OpenAI API error 429 rate limit exceeded

You're hitting the OpenAI API and getting slammed with a 429 error — "Rate limit reached for requests". Your application grinds to a halt, and you're wondering if you've broken something or if OpenAI's blocking you.

OpenAI API error 429 rate limit exceeded

You're hitting the OpenAI API and getting slammed with a 429 error — "Rate limit reached for requests". Your application grinds to a halt, and you're wondering if you've broken something or if OpenAI's blocking you.

What's actually happening

Error 429 means you've exceeded your API rate limits. OpenAI enforces multiple limits simultaneously: requests per minute (RPM), tokens per minute (TPM), and sometimes requests per day (RPD). The exact limits depend on your usage tier and which model you're calling.

When you see `"error": { "message": "Rate limit reached for requests", "type": "requests", "code": "rate_limit_exceeded" }`, you've hit the RPM cap. If the error says `"type": "tokens"`, you've maxed out TPM — common with GPT-4 when processing large prompts or generating long responses. New accounts start at the lowest tier with tight limits: typically 3 RPM and 40,000 TPM for GPT-4, or 3,500 RPM and 200,000 TPM for GPT-3.5-turbo.

Your tier increases automatically as you spend more. After spending $5, you jump to tier 1 (higher limits). After $50, tier 2. Check your current tier and exact limits at platform.openai.com/settings/organization/limits — this dashboard shows real-time usage and where you stand.

How to fix it

1. Check your current tier and usage

Go to platform.openai.com/settings/organization/limits. You'll see a table showing your tier, model-specific RPM/TPM limits, and current usage percentages. If you're consistently hitting 80-100%, you need to either upgrade your tier or implement rate limiting.

2. Implement exponential backoff in your code

When you receive a 429 error, wait before retrying. Start with a 1-second delay, then double it with each subsequent failure (2s, 4s, 8s). Most API libraries support this automatically. In Python with the official OpenAI library, wrap your calls in a retry loop or use `tenacity` for automatic backoff. This prevents hammering the API when you're already over the limit.

3. Batch requests when possible

If you're making multiple independent calls, space them out. Add a delay between requests — even 100-200 milliseconds helps if you're running loops. For processing lists of items, implement a queue system that respects your RPM limit instead of firing all requests simultaneously.

4. Reduce token usage per request

If you're hitting TPM limits specifically, trim your prompts. Remove unnecessary context, use shorter system messages, or limit `max_tokens` in your completion parameters. A 2,000-token prompt with a 1,000-token response consumes 3,000 TPM — run that 14 times in a minute and you've exceeded typical tier 1 limits for GPT-4.

5. Upgrade your usage tier

The fastest fix: spend more to unlock higher tiers. Add credits at platform.openai.com/settings/organization/billing. Once you've crossed a spending threshold ($5, $50, $100, etc.), your limits automatically increase within 24 hours. There's no form to fill out — it happens based on your billing history. For details on other API errors you might encounter, check out our guide to common OpenAI API errors.

If that doesn't work

If you've implemented backoff, reduced usage, and you're still getting 429s constantly, your application might have a bug causing request loops. Check your logs for duplicate calls or infinite retry cycles. Review your code for anywhere you're not catching errors properly.

If you believe your tier should be higher based on spending but isn't, contact OpenAI through platform.openai.com/account/support. Include your organization ID, current spending total, and the specific limits you're seeing. Response times vary — typically 2-5 business days. See our full guide on contacting OpenAI support for what to include.

For billing issues like unexpected charges from excessive API usage, review our guide on understanding unexpected charges.

Questions people actually ask

Q: Will OpenAI raise my limits if I ask nicely?

A: No. Limits increase automatically based on spending. There's no manual review process or approval system for standard API usage. Spend more, wait 24 hours, limits go up.

Q: How long does a 429 error last?

A: Rate limits reset every minute. If you hit your RPM cap at 2:47:30 PM, it resets at 2:48:00 PM. Token limits work the same way — they're rolling minute-by-minute windows.

Q: Can I pay to skip rate limits entirely?

A: Not exactly. Higher tiers have much higher limits, but limits always exist. At tier 5 (after $1,000+ spending), you might get 10,000 RPM for GPT-4 — effectively unlimited for most applications.

Q: Does error 429 affect my billing?

A: No. Failed requests don't consume credits or count toward your usage. You're only charged for successful API calls that return completions.

What to remember

  • Rate limits are per-minute rolling windows — wait 60 seconds and try again
  • Check platform.openai.com/settings/organization/limits for your exact current caps
  • Implement exponential backoff in your code before retrying failed requests
  • Your tier upgrades automatically after spending thresholds — no approval needed
  • Reduce tokens per request if you're hitting TPM limits specifically

---

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

Related help