Lately, I feel that the cursor is not working well. I checked and found that when I select model 3.7 sonnet with agent, it doesn’t work, it gives a not good solution, and I checked the settings and see it using “gpt-4o-mini or cursor-small” instead of premium.
I also found that now Cloud 3.7 doesn’t think at all. It’s unclear what model is being called behind the scenes. If this is to deceive people, such behavior should be prohibited because the output is completely unusable and wastes users’ time.
And it calls too many tool calls
Sneaking to use other models is quite obvious because think 3.7 is relatively slow, while other models are very fast.