Ateron/Gemma-4-Novelist-31B

🖋️ Novelist

"For relaxed minds and calm stories"

📌 Overview

Novelist was a simple merge idea that aimed to write long descriptions, that would not feel too stuttery or flowery. I wanted to see prose that would be rich and at the same time not spill unnecessary water.

🗒️ Merged models

There are models that I used and why:

Component Model	Primary Contribution
nbeerbower/Gemma4-Gutenberg-31B	This is what gives it a voice. Novel-like style that I personally love a lot.
Gryphe/Gemma-4-31B-StyleTune	This, as many people said, brings non-sloppy prose and fresh replies.
bgg1996/Melinoe-Gemma4-31B-VL	This one I found very chaotic on it's own, but with surprising turns. So I used it as spicy material.

⚙️ Mergekit Configuration

Below is the exact mergekit_config.yml recipe used to synthesize this model:

Click to expand Merge Configuration

merge_method: dare_ties
base_model: F:\AI\Merge\Gemma-4-it
tokenizer_source: union
parameters:
  lambda: 1.0
dtype: bfloat16
models:
  - model: F:\AI\Merge\G4-Gutenberg
    parameters:
      density: [0.40, 0.40, 0.40, 0.40, 0.40]
      weight: [0.40, 0.40, 0.40, 0.40, 0.40]

  - model: F:\AI\Merge\StyleTune
    parameters:
      density: [0.35, 0.35, 0.35, 0.35, 0.35]
      weight: [0.40, 0.40, 0.40, 0.40, 0.40]

  - model: F:\AI\Merge\Melinoe
    parameters:
      density: [0.25, 0.25, 0.25, 0.25, 0.25]
      weight: [0.20, 0.20, 0.20, 0.20, 0.20]

🤝 Special Thanks

• Google DeepMind: For providing the base model.
• The Open-Source Community: Creators and all their fine-tuned models that were used in the merge.
• Mergekit Fork: Zerofata - For making mergekit work.
• To Nimbz: This fella for being a smart fella and assisting with advice.