All Products
Search
Document Center

STAROps:Model selection

Last Updated:Sep 18, 2026

Both AI Chat answers and Mission analysis conclusions are generated by a large language model (LLM). Specify the model before you send a question or create a Mission. The model selector on the AI Chat page lets you set the model for each conversation that you start or continue, and for each Mission that you create.

In this topic, the term model refers to the large language model (LLM) that provides inference capabilities. It is not the same concept as the observability model (UModel), which describes the system topology and entity relationships in a workspace. For information about the observability model, see the glossary in Core concepts.

Supported models

In AI Chat, you can switch between the following models in the model selector:

Model

Description

Auto

The default option. The system automatically selects the model for the current inference, so you do not need to specify a model.

Qwen

Models in the Qwen series.

GLM

Models in the Zhipu GLM series.

DeepSeek V4-Pro, DeepSeek V4-Flash

Models in the DeepSeek series.

Enter AI Chat and open the model selector

Follow these steps to enter the AI Chat page and open the model selector. For how to specify the model in each scenario, see Select a model by scenario.

  1. Log on to the Alibaba Cloud console.

  2. The console opens the AI Chat page by default. If you are on another page, click New chat in the left-side navigation pane to return to the AI Chat page.

  3. In the input box area, click the model selector to expand the list of available models.

Select a model by scenario

Start a new conversation

Specify the model before you send the first message, so that the conversation runs on the model that you expect.

  1. Confirm that the Digital Employee and the workspace in the dialog box match the goal of this conversation. The Digital Employee determines the skills and Model Context Protocol (MCP) services that this conversation can call.

  2. Select the model in the model selector, enter your question, and then send it. If you do not select a model manually, the system uses Auto mode by default.

Continue an existing conversation

After you open a previous conversation from the recent conversation list, the system retains the earlier conversation context. You can continue the conversation directly, or use the model selector to specify the model for this interaction.

Create a Mission

When you create a Mission, confirm the model for this task during the intent-alignment step. This is the same entry point where you confirm the Digital Employee and the workspace.

Model switching rules

If the model that you selected goes offline or otherwise becomes temporarily unavailable during a conversation or task execution, the system automatically switches to a different model. It first switches to the latest version of the same model type. If no version is available, the system falls back to Auto mode.

FAQ

How does the system select a model in Auto mode?

In Auto mode, the system uses the model configured on the server side. This mode suits most scenarios and is the default option.

Does switching models affect previous conversation records?

No. Switching models affects only the responses in subsequent conversations. Your conversation history remains unchanged. Different models may answer the same question differently, but switching models does not modify existing conversation content.

Why did the system automatically switch the model that I selected?

This usually happens because the model that you selected became temporarily unavailable during the conversation or task execution. The system then selects a replacement in the order described in Model switching rules.