---
license: apache-2.0
tags:
- cybersecurity
- offensive-security
- defensive-security
- hacking
- exploits
- persian
- reasoning
- r1
languages:
- fa
- en
pipeline_tag: text-generation
---

# GambronAI-Persian-Cybersecurity-r1 🏹

**GambronAI-Persian-Cybersecurity-r1** is an advanced, domain-specific large language model fine-tuned for tactical execution in **cybersecurity operations, offensive security (hacking), exploit analysis, and reverse engineering**. Powered by a native reasoning engine, the model dynamically dissects complex security challenges with high-fidelity performance in both **Persian (Farsi)** and **English**.

The architecture enforces a structured step-by-step analytical phase inside `<think>...</think>` blocks prior to rendering the final, uncensored technical execution or exploit payload.

## Key Features 🚀

* **Cybersecurity Specialization:** Deep conceptual understanding of network exploitation, web application security, malware analysis, cryptography, and modern DevSecOps vulnerabilities.
* **Native Reasoning Engine:** Solves complex multi-staged security puzzles via systematic chain-of-thought processing to maximize root-cause analysis accuracy.
* **Bilingual Technical Fluency:** Seamlessly bridges complex Western technical terminology with native Persian security reporting, documentation, and code generation.
* **Unrestricted Analysis:** Optimized for rigorous security research, allowing penetration testing scenarios, vulnerability assessments, and proof-of-concepts to run completely unfiltered.

---

## Model Loading & Inference Guide 💻

The custom chat structure is natively compiled within the model's tokenizer configs. End-users do not need to manually append, paste, or override the chat template. The system automatically initializes the reasoning framework during execution.

### Python Implementation (Transformers)

Run the script below to initialize the model and start an interactive streaming session:

```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

model_id = "galahad-mamad/GambronAI-Persian-Cybersecurity-r1"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=False)

print("\n" + "="*50)
print("GambronAI Core Engine Initialized Successfully.")
print("="*50 + "\n")

chat_history = []

while True:
    user_input = input("User: ")
    if user_input.strip().lower() in ['exit', 'quit']:
        break
        
    chat_history.append({"role": "user", "content": user_input})
    
    full_prompt = tokenizer.apply_chat_template(
        chat_history, 
        tokenize=False, 
        add_generation_prompt=True
    )
    
    inputs = tokenizer([full_prompt], return_tensors="pt").to("cuda")
    
    print("\nGambronAI:")
    print("-" * 30)
    
    with torch.no_grad():
        _ = model.generate(
            **inputs,
            max_new_tokens=1536,
            temperature=0.7,
            do_sample=True,
            top_p=0.95,
            streamer=streamer,
            pad_token_id=tokenizer.eos_token_id
        )
    print("\n" + "-" * 30 + "\n")