Their models also "became" quite chatty in recent weeks.

Now you spend 2x tokens to achieve the same result because of that.

Sure, you can tell it to "be succinct" or whatever, but the vast majority of people won't so this definitely increased token consumption overall.