scima/Qwen3.8-Whittle-MoE-27B-A17.8B-GGUF-all-quants
LoRA-style vs logic65/Qwen3.8-Whittle-MoE-27B-A17.8B. scima/Qwen3.8-Whittle-MoE-27B-A17.8B-GGUF-all-quants is a GGUF (F16) of different weights. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, shared_expert_gate, v_proj changed, in 64 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/161, MLP 64/320, Attention 64/64