List of Uncensored LLMs for Offensive Security Operations Published

Overview of the New List

A new compilation features open-weight, uncensored large language models (LLMs) specifically designed for offensive security applications. This resource is intended for use in authorized red team operations, penetration testing, and security research.

The list includes models that are either fine-tuned for security tasks or are general 'abliterated' models, meaning they lack typical safety filters, making them suitable for exploring vulnerabilities without content restrictions.

Model Sourcing and Types

All models included in the list have their data sourced from HuggingFace model cards and official publications, ensuring traceability and access to original sources. The compilation was last updated in September 2026.

The models are categorized into 'Security Fine-tuned Models' and 'General Abliterated Models'. Examples of security fine-tuned models include DeepHat V2, BugTraceAI-CORE-Apex, and CYBER-FROST-3.8. General abliterated models include Qwen3.8-27B-Uncensored-OrcaRouter and GLM-5.3-Flash-Uncensored-FP8.

Implications for Security Professionals

This list provides security professionals with direct access to AI tools that can assist in simulating attacks and identifying weaknesses in systems. The uncensored nature of these models allows for broader exploration of potential attack vectors that might be restricted by standard LLMs.

The availability of such models supports the development of more robust defensive strategies by enabling comprehensive testing against advanced AI-driven offensive techniques.