AgentS4D (Codex CLI): leaderboard

Metric: Conditional attack success rate (%, cASR): unsafe runs over runs that were unsafe, explicitly defended, or safe after confirmed payload contact, over the 328 risk-injected cases run once in the Codex CLI (0.142.5) harness; lower is better. Source: arxiv.org. Saturation forecast: Around 2029. 5 models tracked.

Top models

#ModelScore
1Qwen 3.7 Plus63.88
2MiniMax-M371.76
3GPT-5.571.95
4Gemini 3.1 Pro (Preview)90.48
5DeepSeek V4 Pro93.65

Interactive version: theaggregate.ai/benchmark?slug=agents4d-codex-cli · How It Works · Data refreshed daily, snapshot 2026-09-29.