AVIM-AI/gemma-4-E4B-es-latam-mlx
LoRA-style (fewer layers) vs google/gemma-4-E4B-it. AVIM-AI/gemma-4-E4B-es-latam-mlx matches part of google/gemma-4-E4B-it (665 of its 2130 tensors; the rest of google/gemma-4-E4B-it has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, o_proj, q_proj, up_proj changed, in 24 of 42 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/2, Norms 0/278, other 0/127, MLP 72/126, Attention 48/132