When adding multiple messages to the queue, the selected reasoning effort is not assigned to each individual task. Queued messages may all run using the most recently selected level.
It would be useful to add a “Remember reasoning effort for this message” option. Each queued message would store the model and reasoning level selected when it was submitted, for example Low for a simple fix and XHigh for a complex analysis.
This would provide better control over response quality, processing time, and usage without requiring users to manually monitor the queue.