ManniX-ITA/Qwen3.6-27B-Omnimerge-v4
LoRA-style vs Qwen/Qwen3.6-27B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, MLP 0/307, Attention 208/560, lm_head 0/1, Norms 0/325, other 0/1
7 derivatives checked
- ManniX-ITA/Qwen3.6-27B-Omnimerge-v4-GGUF: Same weights, converted
- ManniX-ITA/Qwen3.6-27B-Omnimerge-v4-MTP-GGUF: Same weights, converted
- shaffhausen/Qwen3.6-27B-Omnimerge-v4-GGUF: Same weights, converted
- fribourgquant/Qwen3.6-27B-Omnimerge-v4-MTP-GGUF: Same weights, converted
- shaffhausen/Qwen3.6-27B-Omnimerge-v4-MTP-GGUF: Same weights, converted
- k0valik/Qwen3.6-27B-Omnimerge-v4-IQ4_XS-12.76GiB-GGUF: Same weights, converted
- fribourgquant/Qwen3.6-27B-Omnimerge-v4-GGUF: Same weights, converted
2 other repos hold byte-for-byte identical weight files.