DavidAU/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING
LoRA-style vs coder3101/Qwen3.5-4B-heretic. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, 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 0/4, Norms 0/203, Attention 152/320, MLP 96/196
11 derivatives checked
- mradermacher/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-GGUF: Same weights, converted
- mradermacher/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-i1-GGUF: Quantization
- avan-ag/Qwen3.5-4B-Uncensored-MLX-4bit: Quantization
- xv1rcn/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-MLX-4bit: Quantization
- TomerGamerTV/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-MLX: Quantization
- xv1rcn/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-MLX-8bit: Quantization
- nuofang/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-GGUF: Quantization
- xv1rcn/Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKING-MLX: Partial fine-tune
- DoctorEdoP369/BrimUniverseAi: Full fine-tune
- nightmedia/Qwen3.5-4B-Element: Full fine-tune
- xtools-at/Qwen3.5-4B-Medical-Reasoning: Adapter
2 other repos hold byte-for-byte identical weight files.