Access viggoVet Veterinary AI Models on Hugging Face

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

To access viggoVet Model on Hugging Face, you're required to fill the request form in viggoVet Veterinary AI page Requests are subject to approval.

viggoVet-Reasoning-20B 🩺

🎯 Model Overview

viggoVet-Reasoning-20B is a state-of-the-art veterinary medicine AI model designed specifically for licensed veterinary professionals. Built on a 20 billion parameter architecture, this model provides evidence-based clinical reasoning, diagnostic support, and treatment recommendations across all major veterinary specialties.

What Makes viggoVet Different?

Professional-Grade Reasoning: Advanced chain-of-thought reasoning optimized for clinical decision-making
Veterinary-Specialized: Trained exclusively on veterinary medical literature and clinical cases
No Disclaimers: Designed for professionals who are licensed to practice veterinary medicine
Evidence-Based: Responses grounded in current veterinary medical knowledge and best practices

✨ Key Features

🧠 Advanced Reasoning Capabilities

viggoVet-Reasoning-20B employs sophisticated reasoning mechanisms that mirror clinical thought processes:

Differential Diagnosis Generation: Systematically works through clinical presentations
Risk-Benefit Analysis: Weighs treatment options with clinical nuance
Case-Based Reasoning: Applies knowledge from similar clinical scenarios
Multi-Step Problem Solving: Handles complex cases requiring sequential decision-making

🔬 Clinical Specialties Covered

Emergency & Critical Care: Triage, stabilization, and acute interventions
Internal Medicine: Endocrinology, nephrology, gastroenterology, cardiology
Surgery: Pre-operative planning, surgical techniques, post-operative care
Anesthesiology: Protocol selection, monitoring, pain management
Dermatology: Diagnosis and treatment of skin conditions
Oncology: Cancer diagnosis, staging, treatment protocols
Exotic & Wildlife Medicine: Avian, reptile, small mammal care
Large Animal Medicine: Equine, bovine, small ruminant care
Clinical Pharmacology: Drug selection, dosing, interactions

🎚️ Configurable Reasoning Effort

Adjust the model's reasoning depth based on case complexity:

Low: Quick responses for straightforward cases (~30 tokens reasoning)
Medium: Balanced reasoning for typical cases (~60 tokens reasoning)
High: Deep analysis for complex cases (~120 tokens reasoning)

👥 Target Audience

✅ Intended Users

This model is designed exclusively for:

Licensed Veterinarians (DVM, VMD)
Veterinary Specialists (Diplomates)
Veterinary Residents (under supervision)
Veterinary Faculty & Researchers

❌ NOT Intended For

Pet owners or animal caretakers
Veterinary students (without supervision)
Unlicensed animal care providers
General public seeking pet health advice

🏗️ Model Architecture

Technical Specifications

Specification	Details
Parameters	20 Billion
Architecture	Mixture-of-Experts (MoE) Transformer
Active Parameters	~3.6B per forward pass
Context Length	4,096 tokens (expandable to 32K+)
Precision	BF16 (Brain Float 16)
Vocab Size	100,277 tokens
Attention	Sliding window + Full attention hybrid
Response Format	Harmony format with reasoning tags

Architecture Highlights

Mixture-of-Experts: Activates 4 experts out of 32 per token for efficient inference
Sliding Window Attention: 128-token windows for computational efficiency
Attention Sinks: Maintains long-range context despite sliding windows
Reasoning Tokens: Explicit separation of reasoning from final responses

🎓 Training Details

Dataset

Source: Custom Veterinary Medical Instruct dataset (viggoVet in-house build)
Content: Veterinary clinical cases, treatment protocols, diagnostic workups
Quality: Curated by licensed veterinary professionals
Languages: Primarily English with multilingual medical terminology

💻 Usage

Installation

pip install transformers torch accelerate

Basic Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "viggovet/viggoVet-Reasoning-20B",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

tokenizer = AutoTokenizer.from_pretrained("viggovet/viggoVet-Reasoning-20B")

# Create conversation
messages = [
    {
        "role": "user",
        "content": "10-year-old Golden Retriever with acute vomiting and abdominal pain. What's your diagnostic approach?"
    }
]

# Generate response
inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
    reasoning_effort="medium"  # Options: low, medium, high
).to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=2048,
    temperature=0.2,
    top_p=0.9,
    do_sample=True
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Advanced Usage with Developer Instructions

# Custom system prompt
DEVELOPER_INSTRUCTIONS = """You are viggoVet AI, a veterinary medicine assistant for licensed professionals. 
Provide evidence-based clinical reasoning and recommendations. The user is a licensed veterinarian."""

messages = [
    {"role": "developer", "content": DEVELOPER_INSTRUCTIONS},
    {"role": "user", "content": "Your clinical question here"}
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
    reasoning_effort="high"  # Deep reasoning for complex cases
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=2048)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

🔍 Inference Examples

Example 1: Emergency Case (High Reasoning)

Input:

5-year-old Labrador presenting with acute collapse, pale mucous membranes, 
distended abdomen, heart rate 160. What's your emergency protocol?

Response: (Model provides detailed reasoning)

[Reasoning: This presentation is highly suspicious for hemoabdomen, 
most likely splenic mass rupture given breed and acute onset...]

Immediate Actions:
1. Establish large-bore IV access (18g or 16g catheter)
2. Administer isotonic crystalloids (shock bolus 20-30 mL/kg)
3. Type and crossmatch for blood products
4. Perform AFAST ultrasound to confirm hemoabdomen
5. Stabilize for emergency exploratory laparotomy

[Additional clinical reasoning and detailed protocol...]

Example 2: Diagnostic Workup (Medium Reasoning)

Input:

12-year-old cat with PU/PD, weight loss, poor hair coat. 
What's on your differential list and initial diagnostics?

Response:

[Reasoning: Classic presentation for metabolic disease in geriatric cat...]

Differential Diagnoses (in order of likelihood):
1. Chronic kidney disease
2. Hyperthyroidism  
3. Diabetes mellitus
4. Hepatic lipidosis

Initial Diagnostic Plan:
- Complete blood count (CBC)
- Comprehensive chemistry panel (including T4)
- Urinalysis with specific gravity
- Urine culture if indicated
- Blood pressure measurement

[Continues with detailed reasoning...]

Example 3: Treatment Protocol (Low Reasoning)

Input:

What's your anesthetic protocol for a healthy 2-year-old cat spay?

Response:

Standard Feline Spay Anesthetic Protocol:

Premedication:
- Dexmedetomidine 5-10 mcg/kg IM
- Butorphanol 0.2-0.3 mg/kg IM

Induction:
- Propofol 4-6 mg/kg IV to effect

Maintenance:
- Isoflurane or sevoflurane in oxygen (1.5-2.5%)

Multimodal Analgesia:
- Meloxicam 0.3 mg/kg SQ (once)
- Bupivacaine incisional block

Monitoring:
- HR, RR, SpO2, EtCO2, temperature

⚠️ Limitations

Model Limitations

Not a Replacement for Clinical Judgment
- Recommendations should be evaluated in context of physical exam findings
- Individual patient factors may require protocol modifications
Training Data Cutoff
- Model knowledge reflects training data up to May 2025
- May not include the very latest research or drug approvals
Regional Variations
- Drug availability and protocols may vary by country/region
- Regulatory requirements differ internationally
Rare Conditions
- May have limited knowledge of extremely rare conditions
- Best suited for common-to-moderately-uncommon presentations
Species Limitations
- Strongest in canine and feline medicine
- Good coverage of equine and common exotics
- Limited knowledge of rare exotic species

Technical Limitations

Context Window: 32K tokens (longer cases may need summarization)
Multimodal: Text-only (cannot analyze images, X-rays, etc.)
Real-Time Data: No access to current drug interactions databases
Calculations: May make arithmetic errors in complex dosing

🤝 Ethical Considerations

Professional Use Only

This model is designed to augment, not replace, veterinary professional judgment. It should only be used by individuals who:

Hold valid veterinary licenses
Can critically evaluate AI-generated recommendations
Take full responsibility for patient care decisions
Understand the limitations of AI in medicine

Data Privacy

No PHI: Do not input identifying information about specific cases
De-identification: Anonymize any case details before querying
Local Deployment: Consider on-premises deployment for sensitive cases

Bias Considerations

Training data primarily reflects North American veterinary practice
May have implicit biases toward Western veterinary medicine approaches
Limited representation of traditional or alternative veterinary medicine

No Liability

Model outputs are for informational purposes only
Users assume full professional and legal liability for clinical decisions
Not FDA-approved or certified for medical device use

📚 Citation

If you use viggoVet-Reasoning-20B in your research or clinical practice, please cite:

@misc{viggovet-reasoning-20b,
  title={viggoVet-Reasoning-20B: A Veterinary Medicine Reasoning Model},
  author={ViggoVet Team},
  year={2025},
  howpublished={\url{https://huggingface.co/viggovet/viggoVet-Reasoning-20B}},
}

🙏 Acknowledgments

Base Architecture: Advanced transformer architecture with MoE design
Dataset Contributors: Veterinary professionals who contributed to training data
Open Source Community: Transformers, PyTorch, and related libraries

📧 Contact

For professional inquiries, collaborations, or feedback:

Email: [email protected]
Linkedin: @viggovet
Website: viggo.vet

Built with ❤️ for the veterinary profession

🩺 Empowering veterinarians with AI-assisted clinical reasoning 🩺

Downloads last month: -

Safetensors

Model size

21B params

Tensor type

BF16

Collection including viggovet/viggoVet-Reasoning-20B

Veterinary AI

Collection

LLMs that suit all veterinary medicine professionals • 8 items • Updated 10 days ago