Cupid-Qwen3-4B-v0.1 is my first try at fine-tuning. This is a LoRA fine-tune of Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2 specifically for non-reasoning, uncensored Roleplay (RP).

Nowadays, RP models are either massive or are heavily reasoning-focused, which I found just slow down the experience. The objective of Cupid-Qwen3-4B-v0.1 is to bring back the feeling of the LLama 2 and 3 RP fine-tunes.

Available Formats: This repository includes only the 16-bit PyTorch/Safetensors version, for the GGUF quantizations visit this repo.

Note: This model is based on Qwen3 architecture. Ensure you have the necessary dependencies for Qwen models if you run into issues.