upset with the new models

hi! i use deepseek directly from the API and i loved roleplaying with reasoner until V4 came out. i used to get 9-11 paragraphs of rich responses, but now i, i’d be lucky if i got 4, and this is without using a prompt. i made my own prompt but it still doesn’t give rich responses anymore. deepseek V4 is dirt cheap, and i know that. but that comes at the cost of losing rich, in-depth responses. using R1 on openrouter costs nearly 7 times as more than using the models straight from the API. this is a hobby an