Terminal Bench 2.1

Roybench

Evidence-backed Terminal Bench 2.1 Full88 rankings by provider route, model, effort, and OMP harness.

EnvironmentDetecting…
Generated atNot reported
Data source/data/leaderboards/terminal-bench-2.1-full88.json
Loading leaderboard data…
Source rowsPublished benchmark records
Stable identitiesUnique provider, backend, model, effort, and harness combinations
Best accuracyExact rankable rows
Top pass coverageTasks passed / expected
Scoring and ranking semantics

Ranked evidence

Top performers — Accuracy

Controller score, then version-independent route identity

Full comparison

Benchmark configurations

Terminal Bench 2.1 Full88 rows follow the selected metric ordering
Dataset-level provenance and fields