Skip to main content

​
Quick Answer

Recommended Setting: When creating a token, choose “Pay-per-Use Priority” billing mode, which suits most scenarios.
Although the system provides 5 billing types, using “Pay-per-Use Priority” by default covers all model calling needs. Token Billing Mode Selection Token Billing Mode Selection

​
5 Billing Modes Explained

​
1. Pay-per-Use (Token-based)

Definition: Charges based on the number of Tokens for input and output, pay for what you use. Applicable Models:
  • Text Generation Models: GPT-4, Claude, Gemini, DeepSeek, etc.
  • Multimodal Understanding Models: Models supporting image/audio input
  • Special Image Models: gpt-image-1 (charged by Tokens)
Billing Formula:
Examples:
  • gpt-4o: Input $5/million tokens, Output $15/million tokens
  • claude-3-5-sonnet-20241022: Input $3/million tokens, Output $15/million tokens
gpt-image-1 Special Note: Although it’s an image generation model, it charges by Tokens. Factors affecting Tokens include:
  • Image resolution (1024x1024, 1792x1024, etc.)
  • Image quality (standard, hd)
OpenAI provides a detailed billing table with different Token consumption for different resolutions and qualities.

​
2. Pay-per-Call

Definition: Fixed charge per call, unaffected by input and output Tokens. Applicable Models:
  • Image Generation Models: DALL-E, Flux, Sora Image, etc. (except gpt-image-1)
  • Video Generation Models: Sora Video, VEO, etc.
Billing Formula:
Examples:
  • gemini-3-pro-image (alias nano-banana-pro): $0.09/call
  • sora_video2: $0.15/call (10-second video)
  • flux-1.1-pro: $0.04/call
Advantages of Pay-per-Call:
  • Transparent pricing, fixed cost per generation
  • No need to calculate Token consumption
  • Suitable for fixed output scenarios like images/videos

​
3. Hybrid Billing

Definition: Supports both pay-per-use and pay-per-call billing, automatically selected based on the model. Status: ⚠️ Not Applicable
Currently, API.YI platform does not recommend using “Hybrid Billing” mode, as it may cause billing confusion. Use “Pay-per-Use Priority” instead.
Definition: Smart billing mode that prioritizes pay-per-use billing when a model supports both; automatically switches to pay-per-call if the model only supports that. Why Recommended?
  • ✅ Includes Pay-per-Call: Can call image/video models charged per call
  • ✅ Includes Pay-per-Use: Can call text/multimodal models charged per use
  • ✅ Auto-Adapts: System automatically selects the most appropriate billing method
  • ✅ Covers All Scenarios: All 400+ models supported
Billing Logic:
Example Scenarios:
Recommendation Reason: Using “Pay-per-Use Priority” token allows calling all models without creating different tokens for different billing modes.

​
5. Pay-per-Call Priority

Definition: Prioritizes pay-per-call billing when a model supports both; automatically switches to pay-per-use if the model only supports that. Applicable Scenarios:
  • Scenarios requiring fixed costs
  • Mainly using image/video generation models
Billing Logic:
Usage Suggestion: Unless there’s a clear cost control requirement, use “Pay-per-Use Priority” as text models are usually more cost-effective with pay-per-use billing.

​
How to Choose Billing Mode?

Pay-per-Use Priority (Default Recommended)

Applicable Scenarios:
  • Using text, image, and video models simultaneously
  • Don’t want to create different tokens for different models
  • Need maximum flexibility
Advantages:
  • Covers all 400+ models
  • System automatically selects optimal billing method
  • No additional configuration needed

​
Special Scenarios

Scenario: Only using GPT, Claude, Gemini and other text modelsRecommended Billing Mode: Pay-per-Use Priority or Pay-per-UseReason: Text models all use pay-per-use billing, both modes have the same effect

​
FAQ

gpt-image-1 is OpenAI’s official image generation model. Although it generates images, its billing method is similar to text models, charged by Tokens.Billing Factors:
  • Image resolution (1024x1024 consumes ~5000 tokens, 1792x1024 consumes ~8500 tokens)
  • Image quality (HD quality increases Token consumption)
Solution:
  • Use “Pay-per-Use Priority” or “Pay-per-Use” token
  • Using “Pay-per-Call” token will prevent calling gpt-image-1
Yes, you can modify it. Steps:
  1. Log in to API.YI Token Management Page
  2. Find the corresponding token, click the “Edit” button on the right
  3. Select “Pay-per-Use Priority” in the “Billing Mode” dropdown menu
  4. Save configuration
Note: Changes take effect immediately without affecting existing balance.
Different priorities:Recommendation: Use “Pay-per-Use Priority” in most cases.
Not by itself. Pay-per-use and pay-per-call describe how charges are calculated; they do not indicate whether a route is official-forward or official-reverse, nor do they represent an image quality tier. Check the relevant model page and token group details for the billing modes and routes available to you.For example, the Nano Banana 2.1 pricing page lists both per-request and token-based billing. The Nano Banana integration guide describes dual AIStudio and Vertex channels. So pay-per-call billing alone is not evidence that a request uses an official-reverse route or that its image quality has changed.Different models, routes, or generation parameters may produce different results. When comparing outputs, check the model ID, token group, and generation parameters; to investigate a specific request, provide its request_id.
Won’t fail immediately, but may not be able to call certain models.Example Scenarios:
  • If token is “Pay-per-Call”, calling gpt-4o will fail (because gpt-4o only supports pay-per-use billing)
  • If token is “Pay-per-Use”, calling gemini-3-pro-image may fail (because this model only supports pay-per-call billing)
Solution: Use “Pay-per-Use Priority” to avoid this issue.
Hybrid billing theoretically supports both pay-per-use and pay-per-call, but in practice may cause:
  • Unclear billing logic
  • Difficult cost prediction
  • System compatibility issues
Alternative: Use “Pay-per-Use Priority” to achieve the same effect with better stability and reliability.

​
Summary Recommendations

Best Practice: When creating a token, choose “Pay-per-Use Priority” billing mode to cover all usage scenarios without creating different tokens for different models.