choonok/VetJarvis-4B-Instruct

VISIONConcurrency Cost:1Model Size:4.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Apr 22, 2026License:otherArchitecture:Transformer0.0K Cold

VetJarvis-4B-Instruct is a 4.5 billion parameter domain-specific large language model developed by CHOONOK COMPANY, based on the Qwen3.5 architecture. It is specifically designed for companion animal veterinary knowledge, education, and responsible AI research, excelling in veterinary reasoning tasks. The model demonstrates significantly enhanced reasoning performance, achieving GPT-5.4-mini-class results with its 'thinking' mode enabled. It is primarily intended for non-commercial veterinary research and educational applications.

Loading preview...

VetJarvis-4B-Instruct: Domain-Specific Veterinary LLM

VetJarvis-4B-Instruct, developed by CHOONOK COMPANY, is a 4.5 billion parameter instruction-tuned model built on the Qwen3.5 architecture. It is uniquely specialized for companion animal (canine and feline) veterinary knowledge, education, and AI research. A key differentiator is its enhanced reasoning performance through an improved post-training pipeline, particularly when its thinking mode is activated, which allows it to achieve performance comparable to GPT-5.4-mini.

Key Capabilities & Features

  • Domain Specialization: Highly proficient in veterinary medicine, trained on de-identified clinical data and reference materials.
  • Enhanced Reasoning: thinking mode significantly boosts performance on complex veterinary reasoning tasks.
  • Benchmark Performance: Outperforms similarly sized open local models (e.g., Qwen3.5-4B, Gemma-E4B) by 15-25 percentage points on the Japanese Veterinary National Examination benchmark, narrowing the gap to leading closed models like GPT-5.4-mini.
  • Multilingual Support: Supports both Korean and English languages.
  • Context Length: Features an 8,192-token context window for SFT.

Good For

  • Veterinary Research & Education: Ideal for academic studies and educational purposes in veterinary science.
  • Responsible AI Development: Supports the advancement of safe and responsible AI technologies in the veterinary field.
  • Specialized Knowledge Retrieval: Excels in areas like hematology/oncology, orthopedics, endocrinology, and clinical decision-making within veterinary contexts.

Limitations

It is crucial to note that VetJarvis-4B-Instruct is not a medical device and is not permitted for clinical use to replace professional veterinary judgment. Users must be aware of potential hallucinations and knowledge cutoff limitations (April 2026).