DexopT/Qwen3-4B-Cybersecurity
DexopT/Qwen3-4B-Cybersecurity is a 4 billion parameter language model, fine-tuned from Qwen3-4B-Instruct-2507 by DexopT. Specialized in cybersecurity, it excels across offensive security, penetration testing, vulnerability analysis, and threat intelligence. This model was trained on over 1.28 million cybersecurity-specific samples, making it highly proficient in technical security domains.
Loading preview...
What is DexopT/Qwen3-4B-Cybersecurity?
This model is a 4 billion parameter language model, fine-tuned by DexopT from the Qwen3-4B-Instruct-2507 base model. It is specifically designed and optimized for cybersecurity applications.
Key Capabilities
DexopT/Qwen3-4B-Cybersecurity demonstrates expertise across a broad spectrum of cybersecurity topics, including:
- Offensive Security: SQL injection, XSS, buffer overflows, reverse shells.
- Defensive Security: CVE analysis, incident response, system hardening.
- Network Security: MITM, ARP spoofing, DNS poisoning, DDoS attacks.
- Active Directory: Kerberoasting, Pass-the-Hash, Golden Ticket attacks.
- Malware Analysis: Ransomware, rootkits, persistence techniques.
- Web Security: SSRF, XXE, CSRF, LFI/RFI, deserialization vulnerabilities.
Training Details
The model was trained using Unsloth SFT on a curated dataset of 1,284,369 cybersecurity samples from the DexopT/cyber_heretic dataset. It utilized a context length of 2048 tokens during training.
Important Considerations
- Refusal Behavior: This model retains some refusal behavior from its base Qwen3 model. A "Heretic" version with refusal directions removed is available at DexopT/Qwen3-4B-Cybersecurity-Heretic-16bit.
- Quantized Versions: GGUF quantized versions (Q8_0 + Q4_K_M) are available for use with
llama.cpp, LM Studio, or Ollama at DexopT/Qwen3-4B-Cybersecurity-GGUF.
Should I use this for my use case?
This model is ideal for researchers, security professionals, and developers working on projects that require deep understanding and generation of content related to cybersecurity. Its specialized training makes it highly effective for tasks such as vulnerability explanation, penetration testing methodologies, malware analysis, and threat intelligence. It is intended for educational and research purposes only, and users should exercise responsibility and ensure explicit permission for any testing on systems.