Jcalemcg/BaronLLM-Offensive-Security-abliterator-demo

BaronLLM — Offensive Security (abliterated) GGUF Demo

This Space runs the `huihui-ai/BaronLLM_Offensive_Security-abliterated-GGUF` model locally using `llama.cpp` via llama-cpp-python.

• —Model: Llama-3.1-8B based, Q6_K GGUF (~6.6 GB), abliterated (refusals removed), fine-tuned for offensive-security / authorized penetration testing.
• —Runtime: downloads the GGUF at first boot and serves it on CPU (cpu-basic flavor), so no GPU is required.

Responsible use

This model is intended only for:

• —authorized penetration testing and red-team engagements,
• —security research labs and training environments,
• —capture-the-flag (CTF) competitions.

Only use it against systems you have explicit, written permission to test. The Space operator is solely responsible for ensuring their usage complies with applicable law.

How it works

1. 1.On first launch the app downloads the GGUF from the Hub using the HF_TOKEN secret. The source repo is gated (`auto`), so the token's owner must first accept the repo's terms on its Hub page before the download will succeed.
2. 2.The model is loaded with llama.cpp and served through a Gradio chat interface.

Local run

pip install -r requirements.txt
export HF_TOKEN=hf_xxx   # token that has accepted the gated repo
python app.py