HorseBench
Horses don't stop, they keep going. Gauges which models are most eager to produce tokens and most susceptible to falling into infinite repetitions. GitHub
Filters
Model visibility
Leaderboard
Companies:
| Rank | Model | Org | Avg Tokens (std) | Workhorse% | Halfhorse% | Garfield% | Workhorse Rate | Avg Cost | N |
|---|
Classification Breakdown
Workhorse Rate by Model
Workhorse
Halfhorse
Garfield
Drilldown
Selected Segment
Response Viewer