Is there any way to improve creative writing for Local Models (Qwen)?

A user is seeking methods to enhance the creative writing capabilities of local Qwen models, specifically Qwen 3.8 flash next at q4 with llama.cpp. They note that while Qwen models excel in research and HTML file generation compared to Gemma or Muse, the latter are superior for creative writing. The user is looking for ways to bridge this gap in Qwen's performance.