Channel Grouping

What is Channel Grouping

Channel grouping is a way for the platform to categorize and manage different API channels. Through grouping, the platform can classify channels of different sources, stability, and prices, allowing users to select the appropriate group according to their needs when calling a model.

The main differences between groups are:

• Price: Different groups are charged at different rates
• Stability: Advanced groups typically use more stable upstream channels
• Speed: Some groups may differ in response speed

Group Description

The platform offers multiple groups, each with different pricing and stability. After logging into the platform, you can view all groups and their respective prices on the Model Pricing page.

How to Choose a Group

• Personal Use / Testing: Choose the default group, which has lower prices
• Production / Enterprise Use: It is recommended to select an advanced group for higher stability

Please refer to the Model Pricing page for specific group details.

How to Use a Group

In an API request, specify the group by adding a group suffix after the token:

Token-GroupName

For example, if your token is sk-xxxxxx and you want to use the premium group, fill in:

sk-xxxxxx-premium

If you do not specify a group, the default group will be used.

Viewing Model Prices

After logging in, click "Model Pricing" in the navigation bar to view the specific prices for all models under each group.

Price Calculation Formula

Actual consumed quota = Group Multiplier × Model Multiplier × (Prompt token count + Completion Multiplier × Completion token count) / 500000

Where:

• Group Multiplier: Determined by the selected group, can be viewed on the Model Pricing page
• Model Multiplier: Determined by the model itself, can be viewed on the Model Pricing page
• Completion Multiplier: Varies by model, common values are as follows:
  • GPT-3.5 series: Completion Multiplier is 2
  • GPT-4 series: Completion Multiplier is 2
  • GPT-4o series: Completion Multiplier is 4
  • For other models, please refer to the Model Pricing page