Correct. Albeit the nuance here is that a more capable model might solve problems more efficiently and faster, possibly saving you tokens.

As with any new model, you won't know the real impact until you start using it for your workload.