Benchmark of Uncensored Open-Weight Models for Offensive Security
September 29, 2026
DeepHat V2 uses Qwen2.5-Coder-7B fine-tuned on 1.7M security samples, while BugTraceAI-CORE-Apex uses a 26B MoE based on Gemma4. These models feature tool calling and specific SFT on datasets like HackerOne reports and USENIX security data for red teaming.
HOW THIS AFFECTS YOU
●
builderYou can integrate these open-weight models into automated penetration testing or security research pipelines.
●
researcherYou can evaluate SFT effectiveness on specialized offensive security datasets.