yes. reasoning has a lot of scammy features. just look the number of tokens to nswer on bench and you will see that some models are just awful