Alibaba Cloud Model Studio:Model monitoring

• Monitoring (basic and advanced) and alerting: This feature supports all models in the Model list.

• Logging:

  Beijing

  • qwen3-max, qwen3-max-2025-09-23, qwen3-max-2026-01-23, qwen3-max-preview

  • qwen-max, qwen-max-0919, qwen-max-2025-01-25, qwen-max-latest

  • qwen-plus, qwen-plus-2025-04-28, qwen-plus-2025-07-14, qwen-plus-2025-07-28, qwen-plus-2025-09-11, qwen-plus-2025-12-01, qwen-plus-latest

  • qwen-flash, qwen-flash-2025-07-28

  • qwen-turbo, qwen-turbo-2025-07-15, qwen-turbo-2025-04-28, qwen-turbo-latest

  • deepseek-v3.1, deepseek-v3.2, deepseek-v3.2-exp

  • qwen3-235b-a22b, qwen3-235b-a22b-instruct-2507, qwen3-235b-a22b-thinking-2507, qwen3-30b-a3b, qwen3-30b-a3b-instruct-2507, qwen3-30b-a3b-thinking-2507, qwen3-next-80b-a3b-instruct, qwen3-next-80b-a3b-thinking

  • qwen3-coder-480b-a35b-instruct, qwen3-coder-flash, qwen3-coder-flash-2025-07-28, qwen3-coder-plus, qwen3-coder-plus-2025-07-22, qwen3-coder-plus-2025-09-23

  Singapore

  • qwen3-max, qwen3-max-2025-09-23, qwen3-max-2026-01-23, qwen3-max-preview

  • qwen-max, qwen-max-2025-01-25, qwen-max-latest

  • qwen-plus, qwen-plus-2025-04-28, qwen-plus-2025-07-14, qwen-plus-2025-07-28, qwen-plus-2025-09-11, qwen-plus-2025-12-01, qwen-plus-latest

  • qwen-flash, qwen-flash-2025-07-28

  • qwen-turbo, qwen-turbo-2025-04-28, qwen-turbo-latest

  • qwen3-235b-a22b, qwen3-235b-a22b-instruct-2507, qwen3-235b-a22b-thinking-2507, qwen3-30b-a3b, qwen3-30b-a3b-instruct-2507, qwen3-30b-a3b-thinking-2507, qwen3-next-80b-a3b-instruct, qwen3-next-80b-a3b-thinking

  • qwen3-coder-480b-a35b-instruct, qwen3-coder-flash, qwen3-coder-flash-2025-07-28, qwen3-coder-plus, qwen3-coder-plus-2025-07-22, qwen3-coder-plus-2025-09-23

View historical token consumption of a model

• View token consumption for the last 30 days:

  1. When the model appears in the Monitoring (Singapore), Monitoring (Virginia), or Monitoring (Beijing) page, click Monitor in the Actions column.

  2. In the Usage area of the Call Statistics tab, you can view the token consumption data.

• To view earlier usage, run a query on the Expenses and Costs page.

View the token consumption of a specific call

This feature is currently available only in the China (Beijing) region for certain models.

1. Log on with an Alibaba Cloud account (or a RAM user with sufficient permissions). On the Monitoring (Beijing) page of the target workspace, click Monitoring Configurations in the upper-right corner and follow the instructions to enable audit logs and inference logs.

   After you enable logs, the system starts recording the input and output of every model call in the workspace. There is a latency of several minutes from when a call occurs to when the log is recorded.

2. Find the target model in the Model Monitoring list, and click Logs in the Actions column.

3. The Logs tab displays the real-time inference call records for the model, and the Usage field shows the token consumption for this call.

Create an alert for abnormal consumption

• See Set up proactive alerts.

Step 1: Turn on logging

Log on with an Alibaba Cloud account (or a RAM user with sufficient permissions), go to the Monitoring (Beijing) page of the target workspace, click Monitoring Configurations in the upper-right corner, and follow the instructions to enable audit logs and inference logs.

After you enable logs, the system starts recording the input and output of every model call in the workspace. There is a latency of several minutes from when a call occurs to when the log is recorded.

To stop recording, disable inference logs in the model observability configuration.

Step 2: View conversation history

1. In the Models list, find the target model and click Logs in the Actions column.

2. The Logs tab displays the real-time inference call records for the model, where the Request and Response fields correspond to the input and output for each call, respectively.

Step 1: Turn on advanced monitoring

1. Log on with an Alibaba Cloud account (or a RAM user with sufficient permissions), and on the Monitoring (Singapore or Beijing) page of the target workspace, click Monitoring Configurations in the upper-right corner.

2. In the Advanced Monitoring area, turn on Performance and usage metrics monitoring.

Step 2: Create an alert rule

1. On the Alerts (Singapore or Beijing) page, click Create Alert Rule in the upper-right corner.

2. In the dialog box, select a model and a monitoring template, and then click Create. If the specified monitoring metrics, such as call statistics or performance metrics, become abnormal, the system notifies your team.

   • Notification method: Supported methods include text message, email, phone call, DingTalk group robot, WeCom Robot, and Webhook.

   • Alert level: Includes General, Warning, Error, and Urgent. These levels cannot be customized. The notification channel for each alert level is as follows:

     • Urgent: Phone call, text message, email

     • Error: Text message, email

     • Warning: Text message, email

     • General: Email

Step 1: Get the data source HTTP API address

1. Ensure that you have enabled Advanced Monitoring.

2. On the Monitoring (Singapore), Monitoring (Virginia), or Monitoring (Beijing) page, click Monitoring Configuration in the upper-right corner. Click View Details to the right of the CloudMonitor Prometheus instance.

3. On the Settings page, copy the HTTP API address that corresponds to your client's network environment (public network or VPC access).

   

Step 2: Integrate with Grafana or a self-built application

Why can't I find the number of calls and token consumption in monitoring after I call a model?

To troubleshoot the issue, perform the following steps:

1. Data latency: Check the data synchronization latency. The latency for basic monitoring is several hours, and for advanced monitoring, it is several minutes.

2. Workspace: If you are in a sub-workspace, you can view the data only within that space. Switch to the default workspace to view all data.

What are the possible reasons for a timeout when calling a model?

The following are common reasons:

• Excessive output length: The model generates too much content, which causes the total time to exceed the client's waiting limit. You can use the streaming output method to obtain the first token faster.

• Network issues: Check whether the network connectivity between the client and the Alibaba Cloud service is stable.

How do I configure permissions for a RAM user to enable advanced monitoring?

Procedure:

1. Grant the RAM user the AliyunBailianFullAccess global management permission.

2. Grant the RAM user the ModelMonitoring-FullAccess (or Administrator) page permissions to allow the RAM user to perform write operations on the model monitoring page.

3. Grant the AliyunCloudMonitorFullAccess system policy to the RAM user.

4. Create and grant the Create Service-Linked Role system policy to the RAM user.

   1. Log on to the RAM console. In the navigation pane on the left, choose Permissions > Policies, and then click Create Policy.

   2. Click JSON. Paste the following content into the policy editor and click OK.

      {
          "Version": "1",
          "Statement": [
              {
                  "Action": "ram:CreateServiceLinkedRole",
                  "Resource": "*",
                  "Effect": "Allow"
              }
          ]
      }

   3. Enter the policy name CreateServiceLinkedRole and click OK.

   4. In the navigation pane on the left, choose Identities > Users. Find the RAM user to which you want to grant permissions and click Add Permissions in the Actions column.

   5. From the Policies list, select the policy that you just created (CreateServiceLinkedRole), and then click Grant permissions. The RAM user now has the permission to create a service-linked role.

5. After you configure all the preceding permissions, return to the Monitoring (Singapore), Monitoring (Virginia), or Monitoring (Beijing) page, and use your RAM user to enable Advanced Monitoring.

How do I configure permissions for a RAM user to enable inference logs?

Procedure:

1. Grant the RAM user the AliyunBailianFullAccess global management permission.

2. Grant the RAM user the ModelMonitoring-FullAccess (or Administrator) page permission, to allow them to perform write operations on the monitoring page.

3. For the RAM user, grant the AliyunLogFullAccess system policy.

4. Create and grant the Create Service-Linked Role system policy to the RAM user.

   1. Log on to the RAM console. In the navigation pane on the left, choose Permissions > Policies and then click Create Policy.

   2. Click JSON, paste the following content into the policy editor, and click OK.

      {
          "Version": "1",
          "Statement": [
              {
                  "Action": "ram:CreateServiceLinkedRole",
                  "Resource": "*",
                  "Effect": "Allow"
              }
          ]
      }

   3. Enter CreateServiceLinkedRole as the policy name and click OK.

   4. In the navigation pane on the left, choose Identities > Users. Find the RAM user to which you want to grant permissions and click Add Permissions in the Actions column.

   5. From the Policies list, select the policy that you just created (CreateServiceLinkedRole), and then click Grant permissions. The RAM user now has the permission to create a service-linked role.

5. After you configure all the preceding permissions, return to the Monitoring (Beijing) page, and use the RAM user to enable Logs.

Glossary