Torero96Dev/Cupid-Qwen3-4B-v0.1

Overview

Cupid-Qwen3-4B-v0.1 is a 4 billion parameter model developed by Torero96Dev, representing their initial fine-tuning effort. It is a LoRA fine-tune of the existing Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2 model, built upon the Qwen3 architecture.

Key Differentiator

This model's primary distinction lies in its specific optimization for non-reasoning, uncensored roleplay (RP). The developer observed that many contemporary RP models are either excessively large or overly focused on reasoning, which can hinder the fluidity of the roleplay experience. Cupid-Qwen3-4B-v0.1 was created to recapture the direct and responsive feel of earlier LLama 2 and 3 RP fine-tunes.

Available Formats

The repository provides the model in its 16-bit PyTorch/Safetensors format. Users seeking GGUF quantizations are directed to a separate repository hosted by ToreroDev.

Considerations

Users should ensure they have the necessary dependencies for Qwen models when working with Cupid-Qwen3-4B-v0.1, as it is based on the Qwen3 architecture.

Should I use this for my use case?

• Use this model if:
  • Your primary application is uncensored, non-reasoning focused roleplay.
  • You prioritize a fast and responsive RP experience over complex reasoning capabilities.
  • You are looking for a model that emulates the feel of earlier, dedicated RP fine-tunes.
• Consider alternatives if:
  • Your application requires strong reasoning, mathematical, or coding abilities.
  • You need a model for general-purpose instruction following or factual question answering.