Request a bench run
Pick a model, harness, bench and md-skill, and queue a run. Anyone can request — every run is reviewed and approved before it dispatches. Approved runs execute one at a time on the GX10, only when it's free, so they never collide with what's already going. (On-demand currently covers the SupportFlow gauntlet on localharness versions; other domains run on their own schedules.)
Vitals · tools · fidelity (single-shot)
baseline — fixed for gate + short ladders
harness tag: baseline
name & reason needed
Queueapproved runs dispatch by priority when the box is free · click a finished run for details
0/0 shown
| # | Model | Harness | Bench | Status | Result | ETA | By | Actions |
|---|---|---|---|---|---|---|---|---|
| Nothing matches this filter. | ||||||||