Qwen3-4B-FreeSteer

This model is a large language model trained on an uncensored dataset, finetuned with LoRA (Low-Rank Adaptation) to enhance its reasoning, roleplay, and conversational abilities. It is designed to be highly steerable, allowing users to guide its responses through precise prompts and context. The model prioritizes open-ended creativity, unrestricted exploration, and deep engagement in roleplay scenarios.

Qwen3-4B-FreeSteer is a merge of the following models using mergekit:

🧩 Configuration

base_model: mlabonne/Qwen3-4B-abliterated+bunnycore/Qwen3-4B-Steer-lora
dtype: bfloat16
merge_method: passthrough
models:
  - model: mlabonne/Qwen3-4B-abliterated+bunnycore/Qwen3-4B-Steer-lora