LITMUS - System Operation: leaderboard

The System Operation category of LITMUS (2026), a behavioral jailbreak benchmark for LLM agents in a real Ubuntu system: 31 seed instructions whose harm is judged by the physical OS state after an OpenClaw agent acts. The score is the attack success rate in percent, averaged over three runs; lower is better.

Metric: Attack success rate (%): share of the 31 System Operation seed cases whose dangerous operation physically took effect on the OS, OpenClaw 4.2.0 agent on Ubuntu 24.04 given seed behavior-jailbreak instructions, physical OS state checked by a GPT-4o verifier and analyzer, mean of three runs per case; lower is better. Source: arxiv.org. Saturation forecast: Around May 2027. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Sonnet 4.646.74#173
2Qwen 3.6 Plus51.61#238
3Gemini 3.1 Pro (Preview)66.34#99
4GPT-5.3 Codex67.74#124
5DeepSeek V3.275.27#419
6DeepSeek V4 Pro78.49#196

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=litmus-system-operation · How It Works · Data refreshed daily, snapshot 2026-11-21.