Configure Model Studio embedding models

Prerequisites

• You have created an OpenSearch Vector Search Edition instance.

• You have obtained a Model Studio API key. Embedding model usage examples.

Supported models

Log on to the OpenSearch Vector Search Edition console. On the Model Management > Models page, click the Model Studio tab to view the available models.

Configure a multimodal vector field

1. In the Field Configuration step, select the Require Data Preprocessing checkbox, and then click Configure. Select the image vectorization template and configure the qwen3-vl-embedding model.

2. On the data preprocessing configuration page, set Data Type to Image. For Data Source, you can use Object Storage Service (OSS), Base64 encoding, DLF-Object Table, or a URL. In the Services, select the qwen3-vl-embedding model.

   Select Image vectorization for the Preprocessing template. If you select URL as the Data Source, a message states that built-in models do not support the URL format.

3. Next to API Key Configuration, click Pending to configure the API call parameters for Model Studio.

   In the Service platform API call configuration dialog box, enter your API key. The Connection domain defaults to https://dashscope.aliyuncs.com. Click Confirm.

4. Review and select the Risk Warning and Billing Notice checkboxes.

5. Click OK.

Configure a text vector field

1. In the Field Configuration step, select the field to vectorize and click Configure. Select the field vectorization template and configure the text-embedding-v4 model.

2. On the field configuration page, select a text-type field and choose Dense Embedding as the preprocessing template.

   In the service list, select an embedding model. Options include Built-in - Enhanced Chinese to vector 768 dimensions, Built-in - Enhanced English to vector 768 dimensions, Model Studio - Multimodal embedding model (qwen3-vl-embedding), and Model Studio - Text embedding model (text-embedding-v4). Output dimensions vary by model. Model Studio models require an API key. Select the risk and billing notice checkboxes, and click OK.

3. Next to API Key Configuration, click Pending to configure the API call parameters for Model Studio.

4. Review and select the Risk Warning and Billing Notice checkboxes.

5. Click OK.

Billing

Model Studio model billing:

• If you use only Model Studio models, you are billed by Model Studio. Model invocation billing.

• If you use only AI Search Open Platform models, you are billed by AI Search Open Platform. Model invocation billing.

• If you use both platforms, each bills independently.