Research & benchmarks
Leaderboards and benchmarks used to evaluate agents.
| Product | Version | Stars |
|---|---|---|
| HAL (Holistic Agent Leaderboard)✓Princeton SAgE | — | — |
| SWE-bench✓SWE-bench team | — | 6.0k |
| GAIA✓Meta & Hugging Face researchers | — | — |
| τ-bench✓Sierra | — | 1.5k |
| OSWorld✓unknown | 0.1.16 | 3.2k |