plaw/vulnzap-faraday-mini-1
The plaw/vulnzap-faraday-mini-1 is a 7.6 billion parameter GGUF model, LoRA-finetuned from unsloth/Qwen2.5-Coder-7B-Instruct. Developed by PlawLabs, it specializes in security-oriented code tasks, including vulnerability patching, risk explanation, and CWE classification. This model is optimized for short, code-focused prompts, offering high accuracy in resolving vulnerabilities within code snippets.
Loading preview...
VulnZap Faraday Mini 1: Security-Focused Code Model
VulnZap Faraday Mini 1 is a 7.6 billion parameter model, specifically a 4-bit GGUF LoRA-finetuned version of the unsloth/Qwen2.5-Coder-7B-Instruct base model. Developed by PlawLabs, it has been trained on approximately 363 security-oriented code snippets and patches, making it highly specialized for cybersecurity applications.
Key Capabilities
- Patch Suggestion: Provides fixed versions of vulnerable code blocks.
- Risk Explanation: Describes the nature of vulnerabilities and how patches mitigate them.
- CWE Classification: Identifies Common Weakness Enumeration (CWE) classes directly from raw code.
Training and Performance
The model was fine-tuned using LoRA (r=64, α=128) with Unsloth, utilizing a sequence length of 4,096 tokens over 3 epochs. Manual spot-checks on 50 held-out snippets showed that 100% of suggested patches compile, and 76% fully resolve the vulnerability. It is quantized to Q4_K_M via Unsloth's GGUF exporter.
Good For
This model is ideal for developers and security professionals needing to quickly identify and remediate code vulnerabilities. Its strength lies in short, code-focused prompts, making it suitable for integrating into security analysis workflows. Users should note that patches require review before deployment and the model may underperform on exotic frameworks or complex business-logic flaws beyond its training data.