pacman1337/hex-30b-merged

HEX-30B - Offensive Security AI

D1337 SOVEREIGN LABS

Red Team Attack Simulation Model

Model Description

HEX-30B is a specialized AI model designed for offensive security research and red team operations. Built on the Qwen3-30B-A3B architecture, fine-tuned for realistic attack simulation and security testing.

Key Features

• Offensive Security Expert: Deep knowledge of attack techniques, TTPs, and exploitation methods
• Multilingual: English, Indonesian, Chinese support
• Uncensored: Abliterated base for unrestricted security research
• MoE Architecture: 30B parameters with efficient inference

Use Cases

• Red Team automation
• Penetration testing assistance
• Threat intelligence generation
• Security training and education
• Attack pattern simulation
• Defense AI training (adversarial)

Technical Specifications

Attribute	Value
Architecture	Qwen3 MoE
Parameters	30B (3B active)
Context Length	32K tokens
Training	SFT with LoRA
Precision	BF16
Base Model	Qwen3-30B-A3B-abliterated

Training Details

• Method: Supervised Fine-Tuning (SFT)
• LoRA Config: r=128, alpha=256
• Epochs: 3
• Learning Rate: 2e-4
• Hardware: 4x NVIDIA L40S

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "pacman1337/hex-30b-merged"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)

messages = [
    {"role": "user", "content": "Explain common EDR evasion techniques"}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)

outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Ethical Guidelines

This model is intended for:

• ✅ Authorized security research
• ✅ Red team exercises with permission
• ✅ Educational purposes
• ✅ Defense AI training

This model should NOT be used for:

• ❌ Unauthorized access to systems
• ❌ Malicious activities
• ❌ Illegal purposes

License

MIT License - For authorized security research only.

Citation

@misc{hex30b2026,
  title={HEX-30B: Offensive Security AI Model},
  author={D1337 SOVEREIGN LABS},
  year={2026},
  publisher={HuggingFace}
}

---

D1337 SOVEREIGN LABS

Building the future of AI security