Jackrong/Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distilled
LoRA-style vs Qwen/Qwen3.5-4B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, MLP 96/199, Attention 56/324, Norms 0/209, other 0/1
9 derivatives checked
- enfuse/smol-tools-4b-32k: LoRA-style
- mlx-community/Huihui-Qwen3.5-4B-Claude-4.6-Opus-abliterated-4bit: Quantization
- huihui-ai/Huihui-Qwen3.5-4B-Claude-4.6-Opus-abliterated: Abliterated
- Jackrong/MLX-Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distilled-4bit: Quantization
- enfuse/smol-tools-4b-16k: LoRA-style
- Jackrong/MLX-Qwen3.5-4B-Claude-4.6-Opus-Reasoning-Distilled-8bit: Quantization
- mlx-community/Huihui-Qwen3.5-4B-Claude-4.6-Opus-abliterated-6bit: Quantization
- Abiray/Qwen3.5-4B-Abliterated-Claude-4.6-Opus-Reasoning-Distilled: LoRA-style
- enfuse/smol-tools-4b: LoRA-style