Question 1

How is TPM usage calculated?

Accepted Answer

TPM = requests per minute × average tokens per request (input+output). Headroom is (tier limit − usage) / tier limit × 100%.

Question 2

What does burst tokens mean?

Accepted Answer

Burst tokens = (tier TPM limit / 60) × burst duration in seconds. This shows how many tokens you can send in a short burst without exceeding the per‑minute ceiling.

Question 3

How is queue drain time estimated?

Accepted Answer

Given a queue of N requests and average tokens per request, drain time = (N × avg tokens) / (tier TPM / 60) seconds, assuming full utilization.