Credit billing

Updated at:
Copy as MD

Credit is the unified billing unit for model services in Function Compute. Credit is used exclusively for AI model service invocation — other Function Compute resources such as compute time, memory, and network traffic are billed separately. This document explains how Credit is calculated, priced, and offset through resource packages.

How Credits measure usage

Credit measures your usage of model capabilities provided by Function Compute. You are charged based on the tokens consumed during model invocation, converted to Credit using version-specific coefficients.

Version tiers

Function Compute organizes model capabilities into four version tiers, each containing multiple models with similar functionality and performance. The system automatically routes requests to an available model within your selected version tier. You do not need to manage the specific underlying model being scheduled.

For most use cases, start with the Standard version and adjust based on your task complexity and cost requirements.

Version

Positioning

Use cases

Lite

Fast response, low resource consumption

Simple Q&A, text classification, lightweight information extraction

Economy

Cost-effective, balanced capability

Daily conversations, content summarization, medium-complexity tasks

Standard

Comprehensive capability, covers most business scenarios

Complex reasoning, code generation, multi-turn conversations

Flagship

Strongest model capability

Deep reasoning, long-text analysis, high-precision generation

Billing model

When you invoke model services, the system converts your actual token consumption into Credit usage based on the conversion coefficients for the corresponding version.

Credit value and minimum billing unit

  • 1 Credit = ¥ 0.05 (CNY)

  • Credit is the minimum billing unit. Any amount less than 1 Credit is rounded up to 1 Credit.

Measurement method

The billing formulas are:

  1. Credit usage = Input tokens × Input conversion coefficient + Output tokens × Output conversion coefficient

  2. Cost = Credit usage × ¥ 0.05

Version conversion coefficients

Version

Input token coefficient (per token)

Output token coefficient (per token)

Lite

0.0000192

0.000048

Economy

0.000048

0.000288

Standard

0.000096

0.000384

Flagship

0.000168

0.000672

Different versions have different Token-to-Credit conversion ratios due to varying model capabilities. Generally, the more capable the version, the higher the Credit cost per token.

Billing example

The following example uses the Standard version with 500,000 input tokens and 250,000 output tokens in a single invocation:

  • Input Credits: 500,000 × 0.000096 = 48 Credits

  • Output Credits: 250,000 × 0.000384 = 96 Credits

  • Total: 144 Credits

  • Cost: 144 × ¥ 0.05 = ¥ 7.20

Important billing rules

  • Version-specific conversion — Token-to-Credit conversion ratios vary by version. Refer to the platform for precise calculation values.

  • Resource package validity — Credit resource packages are valid for 1 month from the date of purchase. Unused Credits are cleared upon expiration.

  • Deduction order — When you hold multiple resource packages, the package expiring soonest is deducted first.

  • Deduction scope — Credit resource packages apply to all version types: Lite, Economy, Standard, and Flagship.