Alibaba Cloud Model Studio:Manage workspaces

Create a new workspace

You can create a separate workspace, also known as a sub-workspace, for a new service or scenario. The models, resource costs, and user data within a workspace are completely isolated from other workspaces.

Procedure:

1. Go to the Workspaces (Singapore or Beijing) page.

2. In the upper-right corner, click Add Workspace. In the dialog box that appears, enter a name for the workspace.

3. Click OK.

What to do next: Manage workspace members and permissions

Get a workspace ID

The workspace ID is a unique identifier for a workspace that does not change after it is generated. For example: llm-7c72iiw36kd8****.

• To retrieve the IDs of all workspaces under an Alibaba Cloud account: Go to the Workspaces (Singapore or Beijing) page.

  You must use an Alibaba Cloud account or a RAM user with management permissions. Otherwise, an error is reported when you try to access the page.

• To retrieve the ID of the current workspace: On the Model Studio homepage (Singapore or Beijing), click in the lower-left corner, and then click Workspace Details.

  RAM users without administrative permissions can only view the IDs of the sub-workspaces they have joined.

Switch between workspaces

Workspaces are independent of each other. You cannot view the content of all workspaces at the same time. Instead, you must switch to a specific workspace to view its models, cost information, and user data.

How to switch: On the Model Studio homepage (Singapore or Beijing), click in the lower-left corner. You can then view and switch to different workspaces in the list.

Authorize a sub-workspace to call models

A sub-workspace can invoke a specific model, such as qwen-plus, only after it is granted the required permissions for that model.

Procedure:

1. Go to the Workspaces (Singapore or Beijing) page. Find the sub-workspace that you want to authorize and click Authorization & Throttling Settings.

   

2. Find the desired model and click Edit on the right.

   

3. Enable the required permissions and click Save.

   If the workspace has been authorized before, you do not need to authorize it again.

   

Configure rate limiting for model calling

To ensure fair model calling across different services or scenarios, you can set independent calling rate limits for different sub-workspaces.

• Entry point: On the Workspaces (Singapore or Beijing) page, find the desired sub-workspace and click Authorization & Throttling Settings on the right.

• How it works:

  • Within a workspace, rate limiting is controlled at the model level. For example, for the qwen-plus model, the system counts the total number of requests to qwen-plus from all API keys in the workspace and applies limiting based on this count.

  • Model calling limit is measured in two independent ways: Request Number Limit and Token Limit. Limiting is triggered as soon as either limit is reached.

    For example, assume the request rate limit is 20 requests per minute (RPM) and the token rate limit is 100,000 tokens per minute (TPM). If you send 20 requests to the model within one minute and consume only 1,000 tokens (which is below the 100,000 TPM limit), limiting is still triggered because the number of requests has reached the RPM limit.

• Key limitations:

  • The default workspace does not support limiting settings. The limits for all models in this workspace are the same as the Alibaba Cloud account limits.

  • The total number of calls to a model across all workspaces cannot exceed the Alibaba Cloud account rate limit for that model.

  • Model calling limiting applies to real-time calls, but not to batch inference tasks.

  • You cannot configure limiting for a specific member within a workspace.