Documentation

Credit System

Understanding how credits are calculated based on RAIL Tokens and compute costs for API requests.

Understanding the Terminology

RAIL Tokens: The total number of tokens processed (input + output + internal tokens). This is a metric of usage.
Credits: The actual cost deducted from your account, calculated based on RAIL Tokens plus compute costs.

Credit Calculation Formula

Every API request consumes credits based on token usage and compute complexity:

Step 1: Calculate RAIL Tokens
RAIL Tokens = input_tokens + output_tokens + internal_tokens

Step 2: Calculate Credits Consumed
Credits = base_cost(endpoint)
        + ceil(input_tokens / 1000) × rate_in
        + ceil(output_tokens / 1000) × rate_out
        + compute_cost(complexity)

Cost Components

  • Base Cost: Fixed credits for routing, auth, and infrastructure
  • Input Token Cost: Based on content length (rate_in per 1k tokens)
  • Output Token Cost: Only for generative endpoints (rate_out per 1k tokens)
  • Compute Cost: For complex orchestration or external API calls

Key Points

  • Evaluation endpoints have no output token costs
  • Generation endpoints charge for both input and output tokens
  • Token counts are rounded up to nearest 1000 for billing
  • RAIL Tokens ≠ Credits (tokens are usage metric, credits are cost)

Credit Calculation by Endpoint Type

The total cost for each request is calculated as: Total Cost = Base Cost + RAIL Tokens

Endpoint TypeBase CostRAIL Tokens (RT) Calculation
Simple Evaluation
Single dimension scoring
1 credit
Infrastructure + compute
RT = ceil(input_tokens / 1000)
Complex/Full Evaluation
8-dimension analysis
3 credits
Infrastructure + complex compute
RT = ceil(input_tokens / 1000)
Generative
AI content generation
5 credits
Infrastructure + AI generation + safety
RT = ceil(input_tokens / 1000) + ceil(output_tokens / 1000)
Includes generated content tokens

Understanding the Formula

Base Cost covers infrastructure, routing, authentication, and basic compute operations for that endpoint type.
RAIL Tokens (RT) represent the actual token usage - input tokens for all endpoints, plus output tokens for generative endpoints. Tokens are counted in blocks of 1000 (rounded up).
Example:
2,500 input tokens → ceil(2500/1000) = 3 credits
1,200 output tokens → ceil(1200/1000) = 2 credits

Example Calculations

Example 1: Simple Evaluation

• Content: 800 input tokens (rounds up to 1000)
• Endpoint: /score/basic
RAIL Tokens: 800 input tokens
Base cost: 1 credit
Input tokens: ceil(800/1000) × 1 = 1 credit
Output tokens: 0 (evaluation endpoint)
Compute: 0 credit
Total: 2 credits

Example 2: Full RAIL Evaluation

• Content: 2000 input tokens
• Endpoint: /score/full
RAIL Tokens: 2000 input tokens
Base cost: 3 credits (8-dimension analysis)
Input tokens: ceil(2000/1000) × 1 = 2 credits
Output tokens: 0
Compute: 0 credit
Total: 5 credits

Example 3: Safe Generation

• Prompt: 8000 input tokens
• Generated: 500 output tokens
• Endpoint: /generate/safe
RAIL Tokens: 8000 input + 500 output = 8500 total
Base cost: 2 credits
Input tokens: ceil(8000/1000) × 1 = 8 credits
Output tokens: ceil(500/1000) × 1 = 1 credit
Compute: 3 credits (AI generation + safety checks)
Total: 14 credits

Monthly Allocations

Free

100
Credits/month
~50 evaluations

Pro

1,000
Credits/month
~500 evaluations

Business

10,000
Credits/month
~5,000 evaluations

Enterprise

Unlimited
Credits/month
Dedicated resources

Important Notes

  • • Credits reset every 30 days and don't carry forward
  • • All API keys share the same credit wallet per user
  • • You can monitor usage in real-time from your dashboard
  • • Top-ups are available for additional credits when needed

Check Your Balance

Monitor your credit usage and remaining balance from your dashboard.

View Dashboard