Benchmark history

Compare the scores BenchKit actually recorded—run by run.

Step 1

Choose a modelOnly models found in your results folders appear.

Step 2

Choose a benchmarkDifferent slices and harnesses remain separate runs.

Step 3

Inspect what happenedClick a score for loop, repair and task diagnostics.

Model

Benchmark

Harness

Loop status

Runs

Each row is one recorded benchmark result. No averages or estimates.

Run	Score	Passed	Loop status	Repair	tok/s	Harness

Choose a model to inspect its measured prompt and decode throughput.