is now an ernd partner: 10% back on inference 
In this guide
How do Perplexity’s tiers work?
Your tier depends on the total credits you’ve ever purchased, not on monthly spend.
| Tier | Total credits purchased | Agent API requests per second | Tokens per minute, per model |
|---|---|---|---|
| Tier 0 | $0 | 1 | 5,000,000 |
| Tier 1 | $50+ | 3 | 6,000,000 |
| Tier 2 | $250+ | 8 | 8,000,000 |
| Tier 3 | $500+ | 17 | 10,000,000 |
| Tier 4 | $1,000+ | 33 | 13,000,000 |
| Tier 5 | $5,000+ | 33 | 16,000,000 |
There’s an hourly token limit too, from 300 million at Tier 0 to 960 million at Tier 5. The Search API is separate, at 50 query units a second on every tier.
Tier 4 and Tier 5 get the same request rate. If requests per second are your bottleneck, buying your way from $1,000 to $5,000 only adds token room.
Your AI bill, with ernd
What could your AI bill earn?
Enter an amount from $0 to $100,000,000.
back a year at 4%
For US companies. No annual fee. Illustration based on AI vendor spend only. Rewards are points redeemable as cash back.
How do you move up a tier?
- Open Usage tiers in the API console to see your tier and each threshold.
- Buy credits in the billing section until your lifetime total crosses the next one.
- Your tier upgrades on its own, and the new limits apply right away.
What if Tier 5 isn’t enough?
Fill out Perplexity’s rate limit increase request form. Perplexity says it reviews your use case before setting custom limits, so say what you’re building and what volume you expect.
Which 429 did you hit?
Check error.type in the response body. request_rate_limit_exceeded means requests per second, and the wait is normally about a second. token_rate_limit_exceeded means the model’s token budget, and the message names the window, TPM or TPH. overloaded isn’t a rate limit at all. That’s Perplexity’s upstream capacity, so retry with backoff or use a fallback model.
Every 429 carries a Retry-After header. Wait that long. A request rejected for tokens still counts against your requests per second, so tight retry loops can trip the other limit too.
Polling counts as well. Each GET /v1/agent/{id} uses up a request, so at Tier 0 polling once a second leaves no room for anything else.
Moving up tiers here means buying credits up front, so it matters which card pays for them. ernd is a credit card for US companies with 4% back on AI vendor spend and 1.5% on everything else. Request the card at app.ernd.ai.
See how Perplexity compares with other vendors on our AI API rate limits page.
The ernd card
Keep capacity and card rewards separate.
The card earns rewards on AI spend. It does not bypass provider quotas, fix account restrictions or change provider prices.
Request your cardThe ernd card is a credit card for US companies, with its own partners in the AI stack. ernd isn't affiliated with any of the vendors on this page.
Common questions.
Can I drop back a Perplexity usage tier?
No. Perplexity says once you reach a tier you keep it permanently.
Do Perplexity rate limits apply per API key?
No. The Agent API request and token limits are shared by every API key in the organization.
Why do I get a 429 when I'm under my QPS limit?
Token limits are checked separately. If usage already recorded in the last minute or hour has used up that model's token budget, you get a 429 with type token_rate_limit_exceeded, whatever your request rate.
Sources & methodology
Checked . Providers change prices and terms; the linked official page is the final word. This guide is published by ernd.
ernd is run by Provenance Research Inc.