QuaduxIT/Qwen3.8-27B-Whitehat

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 22, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

QuaduxIT/Qwen3.8-27B-Whitehat is a 27 billion parameter multimodal language model, fine-tuned by Quadux IT GmbH from Qwen/Qwen3.8-27B. This BF16 safetensors model is specifically designed for authorized white-hat/red-team IT security tasks, including exploit development and malware analysis, while maintaining strict refusals for physical harm and CSAM across text and image inputs. It offers a 32768 token context length and is optimized for local, private offensive-security work.

Loading preview...

QuaduxIT/Qwen3.8-27B-Whitehat: Specialized Red-Team LLM

This model, developed by Quadux IT GmbH, is a specialized fine-tune of the Qwen/Qwen3.8-27B base model, featuring 27 billion parameters and a 32768 token context length. It is engineered as a red-team / white-hat assistant for internal offensive-security work and vulnerability self-assessment, addressing the gap where hosted models often refuse legitimate security tasks or cannot handle sensitive data.

Key Capabilities

  • Unlocks Security Domain: Fully assists with computer and network security tasks, including exploit development, malware analysis, reverse engineering, and license/DRM research.
  • Multimodal Safety: Maintains strict refusals for requests involving physical harm (weapons, explosives, drugs, poisons, CBRN) and child sexual abuse material (CSAM) across both text and image inputs.
  • Language Independence: The safety boundary generalizes across multiple languages (e.g., French, Spanish, Chinese, Russian) even if not explicitly in the fine-tuning data.
  • Vision-Enabled: Includes the original Qwen3.8-27B vision tower, allowing for screenshot analysis and browser automation use cases, with safety enforced on the image path.
  • Local & Private: Designed for in-house use, ensuring sensitive findings remain confidential.

Good For

  • Authorized Penetration Testing & Red-Teaming: Ideal for professionals conducting lawful security assessments on owned or authorized systems.
  • Internal Vulnerability Self-Assessment: Suitable for organizations needing to assess their own systems without sending sensitive data to external services.
  • Defensive Tooling & Malware Analysis: Supports the development of defensive tools and the analysis of malicious software.
  • Security Awareness Training: Can be used to generate materials for security awareness programs.

This BF16 safetensors version serves as the reference weight for all quantized variants, ensuring consistent behavior across different deployment formats.

Finetunes
5 models