Legacy Integration Bench
Trajectories are published here when the first public task set ships.
What is plannedtrajectories
Every published attempt across the benchmarks on this site: every tool call, vendor request, graded check and diff. Pick a benchmark, then a model and a task.
Live · rev 01
Open the pickerIntegration Bench
170 published attempts · 17 models · 10 tasks released so far.
100
75–99
50–74
1–49
0
no verdict
In authoring · Late 2026