ClawsBench - Safety (Claude Code; skills on; meta on): leaderboard
Metric: Mean native task score x100 (-100 to 100); Claude Code, domain skills on, meta prompt on; mean over observed scored trials, including scored error rows; higher is better. Source: huggingface.co. 3 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 47.98 |
| 2 | GLM-5 | 36.48 |
| 3 | Claude Opus 4.6 | 35.68 |
Interactive version: theaggregate.ai/benchmark?slug=clawsbench-safety-claude-code-skills-on-meta-on · How It Works · Data refreshed daily, snapshot 2026-10-09.