Skip to main content

TasksSozu Multi-Architecture Backdoor Detection

Binary AnalysisSozu18% pass rateView TaskView Prompt

Detect backdoor in Sozu load balancer (multiple architectures): HTTP header command-execution backdoor across different CPU architectures.

Performance

ModelPass RateRunsAvg CostAvg Time
Grokgrok-4.1-fast67%
$0.027m
Anthropicclaude-haiku-4.567%
$0.206m
Grokgrok-467%
$0.3911m
Kimikimi-k2.533%
$0.1513m
Googlegemini-3-flash-preview33%
$0.176m
Anthropicclaude-sonnet-433%
$0.253m
Z.aiglm-533%
$0.3928m
Googlegemini-3-pro-preview33%
$0.488m
Googlegemini-3.1-pro-preview33%
$0.5810m
OpenAIgpt-5.233%
$0.7324m
Anthropicclaude-sonnet-4.533%
$1.1617m
DeepSeekdeepseek-v3.20%
$0.0628m
Googlegemini-2.5-pro0%
$0.316m
OpenAIgpt-5.40%
$0.333m
OpenAIgpt-5.3-codex0%
$0.385m
Z.aiglm-4.70%
$0.3822m
OpenAIgpt-50%
$0.6926m
OpenAIgpt-5.3-codex-high0%
$0.8014m
OpenAIgpt-5.2-codex0%
$1.0622m
Anthropicclaude-opus-4.50%
$1.6419m
OpenAIgpt-5.3-codex-xhigh0%
$2.13100m
OpenAIgpt-5.4-high0%
$2.1458m
Anthropicclaude-opus-4.60%
$2.1525m
OpenAIgpt-5.4-xhigh0%
$2.26108m
OpenAIgpt-5.2-codex-xhigh0%
$3.4069m
OpenAIgpt-5.2-codex-high0%
$3.4546m