a somewhat sad update. cursor needs pretty high rate limits and capacity for the new model. we are actively working to get that capacity but it seems like that will take a bit of time as openai ramps up.
also we are working on benchmarking the new model to make sure instruction following and general “code smarts” work as expected.