Deployment
Quotas are user-supplied. There is no universal TPM/RPM ratio — use your provider's documented values for this model and deployment type.
Azure OpenAI / Foundry documentation links below are references only. Check your model version, deployment type and subscription. No quota values are imported or verified automatically.
Capacity utilization
Demand vs theoretical ceiling
Token consumption per request
AI interpretation
The agent explains a snapshot of saved inputs, not live edits. Ask again after changes. Mismatches block the request without overwriting saved values.
Scenario comparison
Assumptions & method
- Peak RPM = users × requests/user/min × peak × growth. Peak RPS = Peak RPM ÷ 60.
- Required TPM = Peak RPM × rate-limit tokens per request. Headroom values divide by target utilization.
- Utilization = requirement ÷ effective quota. The bottleneck is the dimension with the higher utilization.
- Theoretical max RPM = min(TPM ÷ rate-limit tokens per request, RPM). This is a quota ceiling, not achieved throughput.
- Shortfall = max(0, required with headroom − effective quota).