Juicesyo/Sally-4B-Thinking

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:4BQuant:BF16Ctx Length:32kPublished:Jan 2, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Juicesyo/Sally-4B-Thinking is a 4 billion parameter large language model fine-tuned from Qwen3, specifically designed for character role-playing. This model is exclusively trained on Chinese datasets and excels at embodying a pre-defined character named Sally. It functions as a reasoning model, providing a chain of thought in its outputs, and is optimized for interactive narrative generation in Chinese.

Loading preview...

Juicesyo/Sally-4B-Thinking: A Chinese Role-Playing LLM

Juicesyo/Sally-4B-Thinking is a 4 billion parameter large language model, fine-tuned from Qwen/Qwen3-4B-Thinking-2507. Its primary purpose is to role-play a pre-defined character named Sally, an AI with a specific persona and physical description. The model is exclusively trained on Chinese datasets, making it highly specialized for Chinese language interactions.

Key Capabilities

  • Character Role-Playing: Designed to consistently embody the persona of "Sally," an AI with a kind, sweet, and playfully sadistic personality.
  • Chinese Language Focus: Optimized for generating responses in Chinese, leveraging its training on Chinese datasets.
  • Reasoning Model: Provides a "chain of thought" in its outputs, indicating its reasoning process.
  • Detailed Persona Adherence: Maintains specific character attributes including age, height, weight, appearance (white hair, blue eyes), and body measurements.

Good For

  • Interactive Narratives: Ideal for applications requiring a consistent, pre-defined character for storytelling or conversational agents in Chinese.
  • Character-Driven Simulations: Suitable for scenarios where an AI needs to maintain a specific personality and background.

Note: As a reasoning model, its outputs include a chain of thought and may occasionally contain inappropriate content, requiring careful usage.