Uncensored and Offensive Security AI Models Benchmark

A GitHub repository curates open-weight, uncensored large language models specifically designed for authorized offensive security operations, including penetration testing and red team activities. The benchmark features five security-focused models—including DeepHat V2, BugTraceAI variants, CYBER-FROST-3.8, and CyberPal 2.0—each fine-tuned on security-specific datasets ranging from bug bounty reports to CVE disclosures and HackerOne data. These models are optimized to generate security tooling, exploit code, and penetration testing scripts while operating under Apache 2.0 or Qwen Community licenses.