Retab uses a credit-based pricing system for AI model usage. Different models have different credit costs based on their capabilities and performance characteristics.

​
Credit price

1 Credit = 0.01$

​
Model Pricing

Model FamilyModel VariantCreditsTier
GPT-4.1nano0.1Micro
mini0.5Small
base2.0Large
Gemini 2.5flash-lite0.1Micro
flash0.5Small
pro2.0Large
o3base5.0Reasoning
Retab routerauto-micro0.1Micro
auto-small0.5Small
auto-large2.0Large

​
Extraction API Pricing

This concerns the following endpoints:

​
Pricing Formula

The total cost for an extract request is calculated as: 
credits/page = preprocessing_cost + (n_consensus × model_credits)

​
Credit Tiers

  • 0.1 credits: Micro models (fastest, most efficient)
  • 0.5 credits: Small models (balanced performance)
  • 2.0 credits: Large models (highest capability)
  • 5.0 credits: Reasoning models (highest tier)
Where:
  • preprocessing_cost:
    • 0 credits: For text-based documents (PDF with text, JSON, CSV, etc.)
    • 0.5 credits: For image-based documents requiring OCR (scanned PDFs, images, etc.)
  • n_consensus: Number of consensus runs (typically 1-5, depending on your accuracy requirements)
  • model_credits: The credit cost of the specific model you’re using (see table above)

​
Examples

Example 1: Text PDF extraction with GPT-4.1-Mini
  • preprocessing_cost: 0 credits (text-based PDF)
  • Model usage: 1 consensus × 0.5 credits = 0.5 credits
  • Total: 0.5 credits
Example 2: Scanned document with Gemini-2.5-Pro (3 consensus)
  • preprocessing_cost: 0.5 credits (image-based document requiring OCR)
  • Model usage: 3 consensus × 2.0 credits = 6.0 credits
  • Total: 6.5 credits
Example 3: JSON extraction with Auto-Micro
  • preprocessing_cost: 0 credits (text-based format)
  • Model usage: 1 consensus × 0.1 credits = 0.1 credits
  • Total: 0.1 credits
Example 4: Scanned invoice with Auto-Micro
  • preprocessing_cost: 0.5 credits (image requiring OCR)
  • Model usage: 1 consensus × 0.1 credits = 0.1 credits
  • Total: 0.6 credits

​
Model Selection Guide

Choose Micro models (0.1 credits) when:
  • You need fast, efficient processing
  • Working with simple extraction tasks
  • Cost efficiency is the primary concern
Choose Small models (0.5 credits) when:
  • You need balanced performance and cost
  • Working with moderate complexity tasks
  • Good balance of speed and capability
Choose Large models (2.0+ credits) when:
  • You need maximum capability and accuracy
  • Working with complex reasoning tasks
  • Quality is more important than cost
Choose Reasoning models (5.0+ credits) when:
  • You need advanced logical reasoning
  • Working with complex problem-solving tasks
  • Maximum intelligence is required

​
Parsing API Pricing

This concerns the following endpoints:

​
Pricing Formula

The total cost for a parse request is calculated as: 
credits/page = model_credits
The Parse API follows the same pricing structure as extraction:
  • 0 credits: For text-based documents
  • model_credits: The credit cost of the specific model you’re using (see table above)

​
Examples

Example 1: PDF parsing with GPT-4.1-Mini
  • Model usage: 0.5 credits
  • Total: 0.5 credits
Example 2: Scanned document with Gemini-flash-lite-2.5
  • Model usage: 0.1 credits
  • Total: 0.1 credits
Example 3: JSON parsing with Auto-Micro
  • Model usage: 0.0 credits
  • Total: 0.0 credits
Example 3: Text parsing with Auto-Small
  • Model usage: 0.0 credits
  • Total: 0.0 credits