Momix-44/gemma-4-31B-it-heretic-v2
Abliterated (fewer layers) vs google/gemma-4-31B-it. Momix-44/gemma-4-31B-it-heretic-v2 matches part of google/gemma-4-31B-it (772 of its 1188 tensors; the rest of google/gemma-4-31B-it has no counterpart). On the shared part: Momix-44/gemma-4-31B-it-heretic-v2 is a GGUF (BF16) of different weights. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 34 of 60 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/361, MLP 34/180, Attention 31/230