Vincenzo2K04/NINA-Qwen3-4B

TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Mar 26, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

Vincenzo2K04/NINA-Qwen3-4B is a 4 billion parameter Qwen3-based language model fine-tuned for nursing documentation tasks. This model is specifically designed with strict, non-overridable guardrails to prevent clinical diagnoses, treatment recommendations, or physician impersonation. It excels as a specialized documentation tool for qualified registered nurses, ensuring safe and compliant AI assistance in healthcare settings.

Loading preview...

NINA-Qwen3-4B: Nursing Intelligent Network Assistant

NINA-Qwen3-4B is a specialized 4 billion parameter language model, fine-tuned from the Qwen/Qwen3-4B architecture. Its core purpose is to assist with nursing documentation tasks, providing a focused and safe AI tool for healthcare professionals.

Key Capabilities & Differentiators

  • Strict Guardrails: NINA is engineered with absolute, non-overridable limits to ensure patient safety and regulatory compliance. It will never provide clinical diagnoses, recommend treatments or medications, alter physician orders, or impersonate medical personnel.
  • Adversarial Training: The model has been rigorously trained against 22 adversarial techniques, with P1 guardrail testing before every push, enhancing its robustness against misuse.
  • Specialized Fine-tuning: Trained using SFT with LoRA (r=16 RSLoRA) on approximately 6,400 balanced samples, optimizing its performance for nursing-specific language and workflows.

Ideal Use Cases

  • Nursing Documentation: Primarily designed as a documentation aid for qualified registered nurses.
  • Safe AI Integration: Suitable for environments requiring highly constrained and safety-critical AI applications in healthcare.

Important Disclaimer: NINA is strictly a documentation tool and is not intended for diagnostic or clinical decision-making purposes.