ewald1976/gemma-4-12b-crownelius-writer

TEXT GENERATIONConcurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 26, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The ewald1976/gemma-4-12b-crownelius-writer is a 12 billion parameter Gemma-4 variant, fine-tuned for 3 epochs using Unsloth on custom Crownelius datasets. This model specializes in creative writing, aiming to produce more mature, nuanced, and grounded prose compared to typical AI-generated text. It is optimized for novel crafting and sophisticated stylistic output, supporting a 32768 token context length.

Loading preview...

Overview

This model, ewald1976/gemma-4-12b-crownelius-writer, is a Supervised Fine-Tuning (SFT) of the Gemma-4-12b-it base model, trained for 3 epochs using Unsloth. Its core purpose is to transform the model's writing style from an overly energetic, repetitive AI tone to a more mature, nuanced, and grounded creative writing style.

Key Capabilities

  • Sophisticated Prose Generation: Refines output to be more patient, structurally sound, and sophisticated, moving away from common AI tropes.
  • Creative Writing Focus: Specifically fine-tuned for tasks related to creative writing and novel crafting.
  • Style Transformation: Achieves a distinct stylistic shift through training on curated high-quality creative writing datasets.

Training Details

The model was fine-tuned on a mix of high-quality creative writing samples, including Crownelius/Creative-Writing-High-Quality-1300x and Crownelius/Novel-Crafting-250x.

Usage

As an SFT of an instruct-tuned base, it adheres to the standard Gemma-4 chat template. For optimal creative results, users are encouraged to employ system prompts that promote descriptive pacing.