Estimate the throughput requirements of a multi-agent workflow using workload assumptions you control. This calculator turns operational inputs into a clear primary estimate plus supporting capacity and planning metrics. Use it for early architecture comparisons, budget reviews, or scenario checks, then replace the defaults with measurements from your own traces and provider terms before making production commitments.
Effective throughput discounts ideal parallel capacity for first-pass completion and service availability, then compares it with entered demand.
This steady-state model does not simulate queue distributions. Validate it with load tests when task duration varies widely or shared rate limits apply.
With 16 slots, 8-second duration, 95% first-pass completion, and 99% availability:
No. This result counts completed tasks or workflows per minute, not model tokens per second.
The raw completion rate is multiplied by the share that completes without another attempt.
Include it when you want end-to-end throughput from arrival to completion.
Uneven task lengths, shared bottlenecks, rate limits, and idle gaps can reduce realized throughput.
Expected demand is higher than the estimated effective completion rate by that amount per minute.
| Variable | Effect |
|---|---|
| Parallel slots | Multiplies simultaneous capacity |
| Duration | Longer time reduces completions per minute |
| Completion rate | Discounts work that needs another attempt |
| Availability | Discounts scheduled downtime |