ValiantLabs/Qwen3.6-27B-Esper4
LoRA-style vs Qwen/Qwen3.6-27B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_qkv, in_proj_z, out_proj, up_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, MLP 192/307, Attention 144/560, lm_head 0/1, Norms 0/325, other 0/1
2 derivatives checked
- mradermacher/Qwen3.6-27B-Esper4-i1-GGUF: Quantization
- mradermacher/Qwen3.6-27B-Esper4-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.