Before Cursor 2.0 we could use thinking or hybrid models for inline chat, which is often handy for a quick question or quick edit while staying focused in the code editor.
Now we can’t use thinking models anymore for inline chat (ctrl+K or cmd+K), even with BYOK.
Almost all models being added are labelled as “thinking” reducing even more the possibilities.
Gemini Flash series are well suited for this task (default inline chat model for Antigravity IDE btw) and we could use our own Gemini API key too in Cursor for inline chat.
Not sure why it got removed, you could add it as a experimental setting and let people who want to use thinking models for inline chat, enable the setting.
I don’t see why you hard lock everyone into a single path, it won’t bother anyone.