When using the new Cursor Agent Window, there is no clear visual indicator to distinguish whether the currently selected model is a thinking or non-thinking variant.
Steps to Reproduce
Open the Cursor Agent Window
Select a thinking model (e.g. Claude 4.6 Sonnet Thinking)
Look at the model name displayed in the UI
Expected Behavior
A brain icon (or equivalent visual indicator) should appear next to the model name to clearly communicate that a thinking model is active — consistent with the behavior in the standard chat interface.
For me it’s even worse. I don’t see thinking indicator, new auto mode switcher, no small model gear button, and nothing is happening if Im hovering mouse on model name. Looks that I have worst possible combination of this. Can you please add me to AB testing group with all features turned on, not off ?
Hi all,
Thanks for the screenshots — this is confusing, and it’s not a misreading on your part.
What’s going on
The compact model control on the Agent bar was simplified on purpose (less chrome when the picker is closed).
When the open list shows two rows with the same name and no clear “thinking” cue, it’s still hard to tell variants apart. That’s a real UX gap, not you.
What you can do today
Use the picker search to narrow to the model family you want (e.g. type part of the model name). That doesn’t always distinguish thinking vs non-thinking when the labels match, but it cuts noise if you have a long list.
Use the keyboard in the picker: arrow up/down to move between the two identical-looking lines and select deliberately once you’ve learned which position maps to the variant you want for that session.
Quick sanity check after picking: if you needed the thinking variant and the reply doesn’t show that extended reasoning style, switch to the other duplicate row and try again (clunky, but reliable when the UI doesn’t label them differently yet).
Update Cursor to the latest version so you pick up any picker changes in newer builds.
