How to Change the Model in a Task?
Adjust the model for subsequent requests while retaining task constraints and validation criteria.
Select Before Sending Next Request
Open the model menu in the task input area, select a model available to your current account, then send the next request. Switching models will not overwrite already completed responses and should not be used as a method to undo previous tool operations.
If the current task is still running, first confirm its status before deciding whether to stop or wait, to avoid issuing conflicting instructions simultaneously.
Specify Clear Goals for the New Round of Analysis
The same task can retain existing context, but key constraints should still be clearly stated. Do not just say "switch model and try again," but specify which part needs improvement.
When comparing two responses, focus on verifying the same set of facts and constraints. For example, request the new round to only review the three conclusions lacking sufficient evidence, rather than re-executing the entire scan; existing valid documents and tool outputs should continue to be used to avoid turning model switching into redundant work.
- Mark the conclusions from the previous result that need to be reviewed.
- Add necessary evidence and constraints that cannot be changed.
- Request a recheck according to the same output standards and indicate corrections needed.
When the Change Does Not Take Effect
Check whether you have entered the Auto path of the free local Agent, or if the target model is still locked. If the context is too long, switching models may not solve the issue; first reduce redundant material. Ultimately, rely on the valid options shown in the input area and the new response results, rather than the model's self-description to confirm actual configuration.