Getting Started

Understanding Credits

Credits are the foundation of LangMart's billing system. This guide explains how credits work, how costs are calculated, and how different models affect your spending.

Credit Basics

  • 1 credit = $1 USD
  • Credits never expire
  • Unused credits remain in your account
  • Credits are non-refundable once purchased

How Costs Are Calculated

Every API request costs a certain number of credits based on:

  1. The model you use - Different models have different prices
  2. Input tokens - The text you send to the model
  3. Output tokens - The text the model generates
  4. Special token types - Some models have additional token categories

Token-to-Credit Conversion

The basic formula for calculating request cost is:

Cost = (Input Tokens x Input Price) + (Output Tokens x Output Price)

For example, if you send 1,000 input tokens and receive 500 output tokens using a model priced at $0.001 per 1K input tokens and $0.002 per 1K output tokens:

Cost = (1,000 x $0.001/1K) + (500 x $0.002/1K)
     = $0.001 + $0.001
     = $0.002 (0.002 credits)

Model Pricing Differences

Different models have dramatically different pricing. Here's a general guide:

Economy Models

Low-cost models suitable for simple tasks:

  • Cost range: $0.0001 - $0.001 per 1K tokens
  • Best for: Simple queries, classification, formatting
  • Examples: Smaller Llama models, Mistral 7B

Standard Models

Balanced cost and capability:

  • Cost range: $0.001 - $0.01 per 1K tokens
  • Best for: General conversation, content generation, code assistance
  • Examples: GPT-3.5-turbo, Claude Instant, Llama 70B

Premium Models

High-capability models for complex tasks:

  • Cost range: $0.01 - $0.15 per 1K tokens
  • Best for: Complex reasoning, code generation, analysis
  • Examples: GPT-4, Claude Opus, Gemini Pro

Frontier Models

Cutting-edge models with the highest capabilities:

  • Cost range: $0.15+ per 1K tokens
  • Best for: Advanced research, complex multi-step reasoning
  • Examples: GPT-4 Turbo, Claude 3 Opus

Special Token Types

Some models include additional token categories that may affect pricing:

Cached Input Tokens

  • Tokens from repeated prompts that can be cached
  • Often priced lower than regular input tokens
  • Reduces costs for repetitive tasks

Reasoning Tokens

  • Internal tokens used by models for complex reasoning
  • Some models charge separately for these
  • Used by models with chain-of-thought capabilities

Self-Funded vs. Organization-Funded

Your funding type affects how credits are managed:

Self-Funded Models

When using self-funded models:

  • Credits are deducted from your personal balance
  • You must maintain minimum balance ($0.10 default)
  • Full visibility into your spending
  • You control when to add credits

Organization-Funded Models

When using organization-funded models:

  • Credits come from organization pool
  • Organization admins set spending limits
  • Usage is tracked per member
  • No personal credit purchase needed

Checking Your Balance

You can check your credit balance in several ways:

  1. Dashboard - View balance in the top navigation bar
  2. Settings page - Detailed balance information
  3. API response headers - Credits remaining returned with each request
  4. Billing page - Full transaction history

Cost Optimization Tips

Choose the Right Model

  • Use economy models for simple tasks
  • Reserve premium models for complex needs
  • Test with cheaper models first

Optimize Token Usage

  • Be concise in your prompts
  • Limit output length when possible
  • Use system prompts efficiently

Monitor Usage

  • Set up cost alerts
  • Review usage analytics regularly
  • Identify expensive patterns

Use Caching

  • Cache responses for repeated queries
  • Use cached input tokens when available
  • Implement request deduplication