AI API Rate Limit Calculator

Enter your provider's tokens-per-minute and requests-per-minute limits, the tokens each call costs, and how often a user triggers one. See how many concurrent users your tier holds, which limit binds first, and how much headroom is left before the 429s start.

🔒 100% in-browser — nothing you enter is uploaded

1. Your provider limits

Illustrative example tiers — fill in your real dashboard numbers below. Limits change often and vary by model, so treat presets as a starting point.
Total input + output tokens per minute.
Separate API calls per minute.
Plan below the ceiling to absorb spikes.

2. Your workload

Prompt + context + system.
Completion length. Counts toward TPM too.
How often one active user calls the API.
Concurrent users supported
0
at your per-user call rate
Sustained requests / min
0
binding ceiling
Binding limit
—
the bottleneck
Tier TPM RPM Req/min (TPM) Req/min (RPM) Binds Users

Estimates, not guarantees. Tokens per request = input + output. Requests/min allowed by TPM = (TPM × utilization) ÷ tokens-per-request. Requests/min allowed by RPM = RPM × utilization. The binding limit is the smaller of the two; concurrent users = binding requests-per-minute ÷ requests-per-user-per-minute. Real ceilings shift with streaming, retries, batching, tokenizer differences between models, and per-model limits. Always confirm current numbers in your provider's dashboard.

Stop renting your ceiling by the token.

Rate limits are the tax you pay for living entirely on someone else's API. QADIR OS runs the models it can on hardware you own — zero TPM, zero RPM, zero 429 — and routes only the overflow to a cloud tier. Own the floor, rent just the peak.

✓ You're on the list. We'll be in touch.

Planning AI API capacity without the guesswork

TPM vs RPM — what is the difference?

TPM (tokens per minute) budgets the total tokens — input and output — your account can process each minute. RPM (requests per minute) budgets how many separate calls you can make, regardless of their size. Providers enforce both simultaneously, so your true ceiling is whichever empties first. Large requests hit the TPM wall; a flood of tiny requests hits the RPM wall.

How many concurrent users can a tier hold?

Take the binding requests-per-minute ceiling and divide by how many requests each active user makes per minute. If TPM allows 150 req/min and RPM allows 500 req/min, TPM binds at 150 — and at one request per user per minute that is 150 concurrent users. Halve the per-user rate and you double the users; double the tokens per request and you roughly halve them.

Why do 429 errors appear below my request limit?

A 429 Too Many Requests fires when you cross either limit. You can sit well under RPM and still get throttled because your requests are large and you exhausted the TPM budget instead. Long completions are the usual culprit — output tokens count toward TPM. The table above shows exactly which limit you are hitting at your workload.

Is my data sent anywhere?

No. This calculator runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.