Raxephion/Krea2-Serendipity-V3
LoRA-style vs krea/Krea-2-Turbo. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down, gate, up, wk, wo, wq, wv changed, in 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/129, other 0/36, MLP 96/105, Attention 160/160