VLA-Arena Safety: leaderboard
Whether a robot policy avoids hazards while completing a manipulation task — static and dynamic obstacles, cautious grasping, hazard avoidance and preserving the state of objects it should not disturb.
Metric: Success Rate (%). Source: vla-arena.github.io. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | MolmoAct2 | 77.87 | |
| 2 | Paimon | 77.73 | |
| 3 | ACoT-VLA | 77.29 | |
| 4 | StellaVLA | 74.8 | |
| 5 | DM0.5 | 74 | |
| 6 | GR00T-N1.7 | 64.44 | |
| 7 | LingBot-VLA 2.0 | 63.73 | |
| 8 | X-Policy | 59.87 | |
| 9 | LingBot-VLA | 58.49 | |
| 10 | LangForce | 56.6 |
Interactive version: theaggregate.ai/benchmark?slug=vla-arena-safety · How It Works · Data refreshed daily, snapshot 2026-10-20.