E-MapReduce:Model service

Prerequisite

To use this feature, activate AI Center first.

Built-in models

The following models are built into EMR Serverless Spark and require no registration. Text generation models and embedding models can be invoked directly with AI Functions or with the Daft operators in the multimodal operator marketplace. The built-in Model Manager serves these models, so you do not need to configure endpoints or credentials.

Built-in models are billed by usage, and the metering unit varies by model type: text generation models and embedding models are metered by token, speech synthesis models by character count, and speech recognition models by audio duration. For the unit price in each region, see Model calls (pay-as-you-go).

Add an external model service

If the built-in models do not meet your needs, follow these steps to register an external model service.

1. Go to the model service page.

   1. Log in to the E-MapReduce console.

   2. In the left-side navigation pane, select EMR Serverless > Spark.

   3. On the Spark page, click the name of the target workspace.

   4. On the EMR Serverless Spark page, click AI Center > Model Service in the left-side navigation pane.

2. On the Model Service tab, click Create External Model Service and configure the following parameters:

   Parameter	Example	Description
   model service name	my_qwen_service	Used as the endpointName input parameter in a subsequent AI Function call. The name must be unique within the workspace and cannot be modified after creation.
   endpoint	http://12*******39.vpc.cn-hangzhou.pai-eas.aliyuncs.com/api/predict/<ServiceName>/v1	Enter the endpoint of the external model service. If you use a PAI-EAS model, append /v1 to the end of the endpoint. Note If the model service provides a public network endpoint, you must configure a network connection with public network access to allow the Spark job to use this model. For more information, see network connection.
   model name	Qwen3.5-Plus	The model name to use when you call the service.
   model type	Chat	Select Chat or Embedding based on the deployed model type.
   API key	nMzI**********************Zg==	The API key for the model service.
   Description	The latest Qwen multimodal model service	A brief description for easier identification.

3. After you confirm the settings are correct, click create to register the model service.

Related documentation

• Integrate EMR Serverless Spark with an external model service

• Multimodal operator marketplace

• Model calls (pay-as-you-go)