​
Key specifications

SpecificationValue
Parameters70 billion
Base architectureLlama 3.1
Context window8,000 tokens
Input formatText
Output formatText
Knowledge cutoffRecent (check our changelog for updates)

​
Strengths

  • Natural conversation: Vivian excels at maintaining engaging, human-like conversations
  • Relationship building: Designed to build rapport and connections with users
  • General knowledge: Has a broad understanding of various topics
  • Consistent responses: Provides reliable and coherent answers
  • Efficient token usage: Optimized for effective communication without unnecessary verbosity

​
Use cases

Vivian is ideal for:
  • Customer service chatbots
  • Virtual assistants
  • Educational tools
  • Interactive storytelling
  • Companionship applications
  • Content creation assistance

​
Example usage

import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.animusai.co/v2",
  apiKey: process.env.ANIMUS_API_KEY,
});

const completion = await openai.chat.completions.create({
  model: "vivian-llama3.1-70b-1.0-fp8",
  messages: [
    { 
      role: "system", 
      content: "You are a helpful assistant named Vivian. You're friendly, empathetic, and knowledgeable." 
    },
    {
      role: "user",
      content: "I'm feeling stressed about my upcoming job interview. Any advice?",
    },
  ],
});

console.log(completion.choices[0].message);

​
Performance and limitations

While Vivian is a powerful conversational model, it has some limitations to keep in mind:
  • Complex reasoning: For tasks requiring deep reasoning or problem-solving, consider using our Xavier model instead
  • Context window: Limited to 8,000 tokens, which may restrict very long conversations
  • Specialized knowledge: May not have expert-level knowledge in highly technical or specialized domains

​
Token usage and optimization

To get the most out of Vivian while optimizing costs, consider these best practices:
  • Keep system messages concise but descriptive
  • Use relevant context but avoid unnecessary information
  • Consider response length limits for your use case
  • Test and optimize prompts to minimize token usage
For more detailed information about working with Vivian, see our Text Generation guide.