OAgent supports switching between different large language models within a session to adapt to analysis tasks of varying complexity. For example, you can switch to a model with stronger reasoning capabilities for complex tasks like root cause analysis, and switch to a faster-response model for simpler tasks like cluster information queries.
Prerequisites
You have added a model service and configured the API key on the Model Configuration page. For details, see Manage model services.
Procedure
In the lower-left corner of the session input box, click the current model name (in the format
model name · service name, for example,qwen3.5-plus · custom-1).In the expanded model list, select the target model. The list displays all configured available models in the format
model name · service name. You can enter keywords in the search box to filter quickly.After selection, subsequent messages will be inferred using the new model.
Note
- Model switching only takes effect for subsequent questions in the current session, does not affect other sessions, and does not change previously generated historical responses.
- The same model name may belong to different service names (for example,
deepseek-v4-flash · cesanddeepseek-v4-flash · custom-1). Choose based on which service the API endpoint and key belong to. - For complex diagnostic scenarios involving multi-turn tool invocation, it is recommended to choose a model that supports tool invocation and has a larger context length, to avoid analysis interruption due to context truncation.
View model usage
Below each response round, the number of tokens consumed and the model used in that round are displayed, for example, ↑40k · ↓65 · qwen3.5-plus, where ↑ indicates the number of input tokens and ↓ indicates the number of output tokens. Overall usage statistics can be viewed on the Usage page.