huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
LoRA-style vs Qwen/Qwen3-4B-Thinking-2507. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, o_proj changed, in 36 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 72/108, Attention 36/144
9 derivatives checked
- mradermacher/Huihui-Qwen3-4B-Thinking-2507-abliterated-i1-GGUF: Quantization
- mradermacher/Huihui-Qwen3-4B-Thinking-2507-abliterated-GGUF: Same weights, converted
- DevQuasar/huihui-ai.Huihui-Qwen3-4B-Thinking-2507-abliterated-GGUF: Same weights, converted
- Aquiles-ai/Athenea-4B-Coding: LoRA-style
- Aquiles-ai/Athenea-4B-Thinking: LoRA-style
- Aquiles-ai/Athenea-4B-Math: LoRA-style
- bunnycore/Qwen3-4B-Max-Ties: Full fine-tune
- Aimin12/Qwen3-4B-Thinking-2507-Distill-Claude-Opus-4.6-Reasoning-Abliterated: LoRA-style
- bralynn/pydevmini1: Full fine-tune
1 other repos hold byte-for-byte identical weight files.