Feature request: Model and reasoning selection for Scheduled Tasks
Please allow users to select and lock the model and reasoning level for each Scheduled Task.
Different tasks have different complexity. Some only need a fast model, while complex workflows require stronger reasoning. I want each scheduled job to consistently run with the model and reasoning level I selected, rather than relying on automatic model selection.
It would also be useful to show the actual model and reasoning level used in each task run.
This control now appears in the Scheduled task interface (on the PC app only), but the selected model and reasoning effort do not appear to be honored.
I reproduced this twice on ChatGPT Plus on 11 October 2026:
- Task 1 was explicitly configured as GPT-5.6 Sol with High reasoning. The completed message showed “Used 6 Instant.”
- Task 2 was a separate diagnostic task explicitly configured as GPT-6 Sol with High reasoning. Its completed message also showed “Used 6 Instant.”
The tasks were configured through the Scheduled interface on Windows, and the actual model used was checked through the completed-message information in the iOS app.
The task editor also does not display the saved model or reasoning setting after creation, so there is no way to verify that those selections persisted until the task runs.
Expected: the scheduled run uses the explicitly selected model and reasoning effort.
Actual: both runs were routed through GPT-6 Instant, including the task configured within the GPT-6 family.
The current documentation says that model and reasoning effort can be selected explicitly.
I have also submitted this through ChatGPT’s built-in bug-reporting form. Can anyone else reproduce it by checking the model shown on a completed scheduled-task message?
