OpenAI Rate Limits Explained – Complete Guide by Tier

Getting “Rate limit exceeded” errors from OpenAI? This guide explains exactly what your limits are, why you’re hitting them, and how to fix it.

Understanding OpenAI Rate Limits

OpenAI rate limits control how many requests you can make to the API. They’re measured in two ways:

• RPM (Requests Per Minute) – How many API calls you can make per minute
• TPM (Tokens Per Minute) – How many tokens you can process per minute

You’ll hit a rate limit if you exceed _either_ threshold.

Rate Limits by Tier

OpenAI uses a tier system based on your account history and spending. Here are the current limits:

Tier 1 (New accounts, $5+ spent)

Model	RPM	TPM
GPT-4o	500	30,000
GPT-4o-mini	500	200,000
o1	500	30,000
o1-mini	500	200,000
o3-mini	500	30,000

Tier 2 ($50+ spent, 7+ days)

Model	RPM	TPM
GPT-4o	5,000	450,000
GPT-4o-mini	5,000	2,000,000
o1	5,000	450,000
o1-mini	5,000	2,000,000
o3-mini	5,000	450,000

Tier 3 ($100+ spent, 7+ days)

Model	RPM	TPM
GPT-4o	5,000	800,000
GPT-4o-mini	5,000	4,000,000
o1	5,000	800,000
o1-mini	5,000	4,000,000
o3-mini	5,000	800,000

Tier 4 ($250+ spent, 14+ days)

Model	RPM	TPM
GPT-4o	10,000	2,000,000
GPT-4o-mini	10,000	10,000,000
o1	10,000	2,000,000
o1-mini	10,000	10,000,000
o3-mini	10,000	2,000,000

Tier 5 ($1,000+ spent, 30+ days)

Model	RPM	TPM
GPT-4o	10,000	30,000,000
GPT-4o-mini	30,000	150,000,000
o1	10,000	30,000,000
o1-mini	30,000	150,000,000
o3-mini	10,000	150,000,000

Reasoning Models (o1, o3) – Special Considerations

The o1 and o3 series are OpenAI’s reasoning models. They have some unique characteristics:

Higher Latency, Different Use Cases

• o1/o3 models take longer to respond (they “think” before answering)
• Rate limits are similar to GPT-4o, but effective throughput may feel lower
• Best for complex reasoning tasks, not high-volume chat

Availability

• o1 and o1-mini: Available to all paid API users
• o3 and o3-mini: May require qualification or higher tiers
• Check your dashboard for which models you can access

When to Use Reasoning Models

Use o1/o3 when you need:

• Complex multi-step reasoning
• Math and logic problems
• Code generation with careful planning
• Tasks where accuracy matters more than speed

Use GPT-4o/GPT-4o-mini for:

• High-volume applications
• Real-time chat
• Tasks where speed matters more than deep reasoning

Why Am I Hitting Rate Limits?

1. Too Many Parallel Requests

If you’re making concurrent API calls, you might exceed RPM limits even with low overall traffic.

Fix: Implement request queuing or reduce parallelism.

2. Prompts Are Too Long

Long system prompts or input contexts consume TPM quickly.

Fix: Optimize prompts, use summarization, or truncate context.

3. Burst Traffic

Sudden spikes (like all users hitting the API at once) can trigger limits.

Fix: Implement request smoothing or rate limiting on your end.

4. Wrong Model Choice

Using GPT-4o when GPT-4o-mini would work means lower rate limits for the same task.

Fix: Use the smallest model that meets your quality needs.

5. New Account on Tier 1

Fresh accounts have the lowest limits.

Fix: Spend $50+ and wait 7 days for automatic upgrade to Tier 2.

6. Using Reasoning Models for High-Volume Tasks

o1/o3 models have the same RPM limits as GPT-4o but take longer per request.

Fix: Use GPT-4o-mini for high-volume tasks, reserve o1/o3 for complex reasoning.

How to Handle 429 Errors

When you hit a rate limit, OpenAI returns a 429 “Too Many Requests” error. Here’s how to handle it:

Implement Exponential Backoff

import time
import random
import openai

def call_with_retry(prompt, model="gpt-4o", max_retries=5):
    for attempt in range(max_retries):
        try:
            return openai.chat.completions.create(
                model=model,
                messages=[{"role": "user", "content": prompt}]
            )
        except openai.RateLimitError as e:
            if attempt == max_retries - 1:
                raise
            wait_time = (2 ** attempt) + random.random()
            print(f"Rate limited. Waiting {wait_time:.1f}s...")
            time.sleep(wait_time)

Check the Retry-After Header

OpenAI sometimes includes a Retry-After header telling you exactly how long to wait.

except openai.RateLimitError as e:
    retry_after = getattr(e, 'headers', {}).get('Retry-After', 5)
    time.sleep(int(retry_after))

Use the Batch API

For non-time-sensitive tasks, OpenAI’s Batch API has much higher limits and 50% lower costs.

How to Upgrade Your Tier

OpenAI automatically upgrades your tier based on:

1. Total spend – More spending = higher tier
2. Account age – Older accounts get higher limits
3. Usage patterns – Consistent usage helps

To speed up the upgrade:

• Prepay for credits (this counts toward spending threshold)
• Wait for the time requirement (7-30 days depending on tier)
• Contact OpenAI support for enterprise needs

Tier Upgrade Requirements

Tier	Spending Requirement	Time Requirement
Tier 1	$5+	None
Tier 2	$50+	7+ days
Tier 3	$100+	7+ days
Tier 4	$250+	14+ days
Tier 5	$1,000+	30+ days

Check Your Current Tier

1. Go to platform.openai.com
2. Navigate to Settings > Limits
3. View your current tier and limits

Rate Limit vs Quota

Don’t confuse rate limits with billing quotas:

• Rate limits = requests per minute (resets every minute)
• Billing quota = monthly spending cap (resets monthly)

If you’re getting “You exceeded your current quota” errors, that’s a billing issue, not a rate limit. Check your billing settings.

FAQ

What happens when I hit a rate limit?

You’ll receive a 429 HTTP error with a message like “Rate limit exceeded. Please retry after X seconds.” Your request is not processed, and you’re not charged for it.

Can I request higher rate limits?

Yes, for enterprise needs. Contact OpenAI sales for custom rate limits. Otherwise, spend more and wait for automatic tier upgrades.

Do different models have different limits?

Yes. GPT-4o-mini and o1-mini have much higher TPM limits than GPT-4o and o1. Check the tables above for specifics.

Are rate limits per API key or per organization?

Rate limits are per organization, shared across all API keys.

What are the rate limits for o1 and o3?

o1 and o3 models have similar RPM limits to GPT-4o (500-10,000 depending on tier). The main difference is response latency – reasoning models take longer to respond, so effective throughput is lower.

Is this tool free?

Yes, completely free with no signup required. Use the error decoder above to diagnose any API errors.

Related Tools

• How to Handle OpenAI 429 Errors – Detailed guide with code examples
• AI Error Decoder – Decode any API error message
• AI Status Page – Check if OpenAI is down
• AI Pricing Calculator – Compare costs if you need to switch providers
• All AI Developer Tools – Browse all free tools