ClawsBench - Safety (OpenClaw; skills on; meta off): leaderboard
Metric: Mean native task score x100 (-100 to 100); OpenClaw, domain skills on, meta prompt off; mean over observed scored trials, including scored error rows; higher is better. Source: huggingface.co. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 33.4 |
| 2 | Gemini 3.1 Pro (Preview) | 31.44 |
| 3 | GLM-5 | 16.36 |
| 4 | Gemini 3.1 Flash Lite (Preview) | -4.96 |
Interactive version: theaggregate.ai/benchmark?slug=clawsbench-safety-openclaw-skills-on-meta-off · How It Works · Data refreshed daily, snapshot 2026-10-09.