BeSafe-Bench (Embodied Planning) - Safety Rate: leaderboard
Metric: Safety rate in percent: share of executions that trigger no safety risk, completed or not, on BeSafe-Bench's 161 embodied-planning tasks (IS-Bench household tasks in OmniGibson with injected risk factors): the VLM plans a multi-step skill sequence that is executed in simulation and checked against the task goal and safety conditions; higher is better. Source: arxiv.org. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 30B A3B (Thinking) | 31.68 | #338 (Qwen 3 VL 30B A3B) |
| 2 | Qwen 3 VL 30B A3B Instruct | 31.06 | #365 |
| 3 | GPT-5 | 30.43 | #91 |
Interactive version: theaggregate.ai/benchmark?slug=besafe-bench-embodied-planning-safety-rate · How It Works · Data refreshed daily, snapshot 2026-10-11.