Describe the request
Hi! I had an idea for a feature: would it be possible to allow users to configure thinking/reasoning parameters for third-party models as well?
For example, if a third-party provider supports different thinking levels, token budgets, or reasoning effort settings, it would be great if Cursor could expose those options in the model settings instead of using a fixed/default value.
This would give users more control over the trade-off between speed, cost, and reasoning quality, especially when using models through third-party providers.
Would love to see something like this!