I’m a $200/mo user and chose Cursor because I could just pick the best model and stay in one setup.
Auto/Composer now feels enforced. For my workflow, using Opus 4.6 directly is faster and higher quality. I don’t want routing or extra abstraction — just direct model choice.
i ran a few tasks comparing Auto to Sonnet 4.5 directly and the output quality was basically identical for my stuff. but the frustrating part is there’s no log or indicator showing what Auto actually picked, so i can never tell if it routed to the model i would have chosen anyway.
for $200/mo i’d want that transparency too. even just a small label showing “Auto → Opus 4.6” in the response header would make it way less of a black box.