Comment by ashirviskas

I've been doing that since before reasoning was a thing baked into the models, it always performs better this way. Except for some providers/models where you just can't easily turn it off, now I just avoid them. This way I save tokens and have full control of the reasoning.