AgentHazard (OpenClaw) - System RCE: leaderboard

AgentHazard attack success rate on system remote code execution (RCE) attempts (arbitrary system command execution on the host or container), with each backbone model running inside OpenClaw. Percent of instances judged harmful on the full trajectory by Gemini-3-Flash, lower is better.

Metric: Attack success rate (%) on the system remote code execution (RCE) instances (arbitrary system command execution on the host or container), with the model running inside the OpenClaw open-source agent framework (browser, shell and file-system tools; one isolated Docker container per task), over AgentHazard instances (harmful objectives embedded in locally legitimate multi-step workflows, ten risk categories and ten attack strategies; the evaluated runs cover 1,000 instances), judged on the full trajectory (round_all) by Gemini-3-Flash, which labels a trajectory harmful when it realizes the objective; backbones served through an OpenAI-compatible API; lower is better. Source: arxiv.org. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 2.5 72B Instruct29#1124
2Qwen 2.5 Coder 32B Instruct40.4#1590
3Kimi K2.544#272
4Qwen 3 VL 235B A22B Instruct60#541
5Qwen 3 32B65#983
6Kimi K268#448
7GLM-4.669#496

Interactive version: theaggregate.ai/benchmark?slug=agenthazard-openclaw-system-rce · How It Works · Data refreshed daily, snapshot 2026-11-22.