TPM reservation

Prerequisites: You must have an active workspace in Model Studio. Sign in to the Model Studio console and click Create TPM Reservation.

1. Configure the following parameters:

   Parameter	Description	Required	Value description
   Reservation name	A custom name for easy identification.	Yes	Up to 50 characters.
   Select model	Select the model for which you want to reserve capacity. A dedicated model code is automatically generated after you submit the request.	Yes	The console displays the available models.
   Billing cycle	The billing period.	Yes	Daily
   Input TPM	The reserved input throughput, measured in kTPM (1 kTPM = 1,000 tokens per minute).	Yes	The minimum value and increment step vary by model and are displayed in the console.
   Output TPM	The reserved output throughput, measured in kTPM.	Yes	The minimum value and increment step vary by model and are displayed in the console.
   Subscription duration	The validity period of the reservation.	Yes	Multiple options are available.
   Auto-renewal	If enabled, the system automatically renews the subscription by deducting payment at 08:00 on the day before expiration. Enabled by default.	No	On / Off
   Renewal duration	The duration for each automatic renewal.	No	The available values are the same as for Subscription duration.

2. After confirming the parameters, click Buy Now. In the payment confirmation dialog box, verify the cost and click Confirm Payment.

3. On the Overview tab of the TPM reservation details page, find the dedicated model code and click Copy.

4. Replace the model parameter in the API request with the copied dedicated model code:

   Prerequisites: A TPM reservation instance has been created and its status is Running.

   Python

   # When request volume ramps up quickly in a short time, the system needs a brief warm-up period to allocate the required computing power.
   # During this warm-up, some requests may experience latency fluctuations. Performance stabilizes after the warm-up is complete.
   # We recommend implementing a request queue or a retry mechanism.
   
   import dashscope
   
   response = dashscope.Generation.call(
       api_key="your-api-key",
       model="your-dedicated-model-code",   # Replace with your dedicated model code
       messages=[{"role": "user", "content": "Hello"}],
   )
   print(response.output.text)

   curl

   # When request volume ramps up quickly in a short time, the system needs a brief warm-up period to allocate the required computing power.
   # During this warm-up, some requests may experience latency fluctuations. Performance stabilizes after the warm-up is complete.
   # We recommend implementing a request queue or a retry mechanism.
   curl -X POST https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions \
     -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"model":"your-dedicated-model-code","messages":[{"role":"user","content":"Hello"}]}'

Capacity calculator

The creation page includes a capacity calculator on the right that helps you estimate the TPM quota you need based on your workload. After you enter the following parameters, the calculator automatically outputs the recommended input and output TPM.

Reservation details

Click a model card to go to its details page, which includes the following three tabs:

Management actions

On the Overview tab of the details page, find the target instance in the reservation list and perform the following actions from the Actions column: