icedsoylatte/wz-qwen25-3b-coser-roleplay-sft-v4

TEXT GENERATIONConcurrency Cost:1Model Size:3.1BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jul 1, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The icedsoylatte/wz-qwen25-3b-coser-roleplay-sft-v4 is a 3.1 billion parameter Qwen2-based causal language model developed by icedsoylatte. This model is specifically fine-tuned for roleplay applications, building upon the icedsoylatte/wz-qwen25-3b-chai-roleplay-sft-v4 base. It was trained using Unsloth and Huggingface's TRL library, enabling 2x faster fine-tuning. With a 32768 token context length, it is optimized for engaging and extended conversational roleplay scenarios.

Loading preview...

Model Overview

The icedsoylatte/wz-qwen25-3b-coser-roleplay-sft-v4 is a 3.1 billion parameter language model developed by icedsoylatte, based on the Qwen2 architecture. This iteration is a fine-tuned version of the icedsoylatte/wz-qwen25-3b-chai-roleplay-sft-v4 model, specifically enhanced for roleplay applications.

Key Characteristics

  • Architecture: Qwen2-based, a powerful causal language model family.
  • Parameter Count: 3.1 billion parameters, offering a balance between performance and computational efficiency.
  • Context Length: Supports a substantial 32768 tokens, suitable for maintaining long and complex conversational threads in roleplay.
  • Training Efficiency: Fine-tuned using Unsloth and Huggingface's TRL library, which facilitated a 2x speedup in the training process.

Primary Use Case

This model is explicitly designed and optimized for roleplay scenarios. Its fine-tuning focuses on generating coherent, engaging, and contextually appropriate responses for interactive storytelling and character simulation. Developers looking for a specialized model for creating immersive roleplay experiences will find this model particularly suitable.