FEATUREBENCH
FeatureBench
Leaderboard
| Rank | Model | Scaffold | %Passed | %Resolved | Harness Org | Model Org | Model Release | |
|---|---|---|---|---|---|---|---|---|
| Rank | Model | Scaffold | Effort | %Passed | %Resolved | Harness Org | Model Org | Model Release |
1 Routing: Routing mode may route operations to different models, even if a specific model is selected.
2 Mock: Tool calls are string-encoded and simulated.
Evaluation settings: Unlimited cost · Timeout: 3600s · Max steps: unlimited