abhaybhargav/PWNISMS-Threat-Model-Structured

TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kPublished:Apr 27, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

abhaybhargav/PWNISMS-Threat-Model-Structured is a 1.5 billion parameter Qwen2.5-Instruct derivative model fine-tuned to generate structured threat models. It specializes in emitting valid JSON outputs conforming to a PWNISMS threat model schema, covering seven security domains with concrete mitigations. This model is optimized for assisting security architects in drafting structured threat models from system descriptions.

Loading preview...

PWNISMS-Threat-Model-Structured Overview

This model, developed by abhaybhargav, is a 1.5 billion parameter variant of Qwen2.5-1.5B-Instruct specifically fine-tuned to produce structured threat models in valid JSON format. It leverages LoRA fine-tuning on MLX and is available in both MLX and GGUF releases, including quantized versions for broader local compatibility.

Key Capabilities

  • Structured JSON Output: Designed to emit JSON exclusively, adhering to a predefined threat_model_schema.json for consistency.
  • PWNISMS Framework: Generates threat models across seven critical domains: Product, Workload, Network, IAM, Secrets, Monitoring, and SupplyChain.
  • Concrete Mitigations: Outputs include specific, actionable mitigations referencing technologies, configurations, or processes.
  • Chat-based Interaction: Expects a system prompt defining the architect's role and requirements, followed by a user prompt with a markdown system description.
  • Output Validation: Internal evaluations show 16/20 outputs parse as JSON, and 12/20 pass full Pydantic validation and cover all seven domains, though human review is always recommended.

Good For

  • Security Architects: Assisting in the rapid drafting of structured threat models.
  • Automated Threat Modeling: Generating initial threat model drafts that can be programmatically validated against a schema.
  • Integration with Security Workflows: Providing structured data for further analysis, audit, or compliance processes.

Limitations: The model can require up to 12,000 output tokens for long scenarios, and lower token caps may truncate JSON. It is not a substitute for expert human review in production assurance.