Skip to main content
Understanding how tokens work and how billing is calculated in mixus.

What Are Tokens?

Tokens are the fundamental unit of measurement for AI model usage in mixus. Every interaction with an AI model consumes tokens based on:
  • Input tokens: The text you send to the model
  • Output tokens: The response generated by the model
  • Processing complexity: Some operations require additional computational resources

Key Concepts

Token Counting

  • Tokens are roughly equivalent to words, but vary by language and model
  • English text averages ~1.3 tokens per word
  • Code and structured data may have different token ratios
  • Each model has its own tokenization method

Billing Structure

mixus uses a transparent, usage-based billing model:
  • Pay-per-token: Only pay for what you use
  • No hidden fees: Clear pricing for all features
  • Monthly billing: Consolidated invoices
  • Usage tracking: Real-time monitoring

Token Types

Input Tokens

Tokens from your prompts, uploaded files, and context:
  • Chat messages
  • File content
  • System prompts
  • Memory context

Output Tokens

Tokens generated by the AI model:
  • Chat responses
  • Generated code
  • Analysis results
  • Tool outputs

Processing Tokens

Additional tokens for special operations:
  • Web search queries
  • Document analysis
  • Image processing
  • Agent execution

Pricing Tiers

Free Tier

  • 1,000 tokens per month
  • Access to basic models
  • Standard support

Pro Tier

  • $20/month base
  • 100,000 included tokens
  • Premium model access
  • Priority support

Team Tier

  • $50/month base
  • 500,000 included tokens
  • Advanced collaboration features
  • Team management tools

Enterprise

  • Custom pricing
  • Unlimited tokens
  • Dedicated support
  • Custom integrations

Cost Optimization

Best Practices

  1. Use appropriate models: Choose the right model for your task
  2. Optimize prompts: Shorter, clearer prompts use fewer tokens
  3. Manage context: Remove unnecessary conversation history
  4. Batch operations: Group similar tasks together

Model Efficiency

  • GPT-4o mini: Most cost-effective for simple tasks
  • GPT-4o: Best balance of capability and cost
  • Claude 3.5 Sonnet: Excellent for analysis and reasoning
  • o1-preview: Use for complex problem-solving only

Next Steps

I