Agent runtime comparator

Measure what the harness adds to the model.

Compare two agent setups across quality, latency, and cost to isolate the harness effect.

Try it now

See what the wrapper changes

Compare two agent setups with a practical weighting: task success first, then latency and run cost.

Harness AA success %A latency (sec)A cost / run ($)

Harness BB success %B latency (sec)B cost / run ($)

Live result

A 11.4-point harness effect under this weighting.

Harness A88.4weighted score

Harness B99.8weighted score

Difference11.4points

Frontier Tools