Mykee/HOTHUN-Stheno-3.2-v1.1

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:8kPublished:Apr 2, 2026License:cc-by-nc-4.0Architecture:Transformer Open Weights Cold

Mykee/HOTHUN-Stheno-3.2-v1.1 is an 8 billion parameter language model developed by Mykee, fine-tuned for NSFW chat in Hungarian while preserving the original model's style. It is optimized for mobile deployment with a focus on Q4_0 quantization and a context length of 8192 tokens. This model excels at generating vulgar and explicit Hungarian text, making it suitable for personal NSFW chat applications.

Loading preview...

HOTHUN-Stheno-3.2-v1.1 Overview

Mykee/HOTHUN-Stheno-3.2-v1.1 is an 8 billion parameter model specifically fine-tuned for NSFW chat in Hungarian. Developed by Mykee, this project aims to adapt existing models with some Hungarian language capabilities into NSFW-proficient chatbots, while largely retaining their original style. A key design constraint was mobile compatibility, limiting the base model size to 8-9B parameters and optimizing for Q4_0 quantization.

Key Capabilities

  • Hungarian NSFW Chat: Specialized in generating vulgar and explicit content in Hungarian.
  • Style Preservation: Designed to maintain the original base model's thought process and style, but in Hungarian.
  • Mobile Optimization: Built with mobile deployment in mind, ensuring performance on devices supporting Q4_0 quantization.
  • Adaptive Generation: Supports both Mirostat and traditional parameter settings, allowing users to switch between modes to prevent looping or enhance story coherence.

Unique Aspects

This model addresses significant challenges in Hungarian NLP, including cleaning noisy datasets, expanding NSFW vocabulary, and refining complex Hungarian conjugations. It was trained using a mix of public Hungarian datasets and Gemini-generated NSFW and grammatical correction datasets. The model's strong vulgar language makes it primarily suitable for personal use.

Recommended Usage

Users should be mindful of the model's sensitivity to Hungarian text, especially in system prompts and character sheets. Experimenting with generation parameters is crucial; specific settings are provided for both Mirostat and traditional modes to achieve optimal responses. If initial responses are unsatisfactory, regenerating the output is often effective.