etemiz/Ostrich-27B-Qwen3.6-260526

Full fine-tune vs Qwen/Qwen3.6-27B. Fine-tune: 100% of the weights changed, across 64 of 64 layers.

Tensors changed: Embeddings 5/5, MLP 307/307, Attention 560/560, lm_head 1/1, Norms 325/325, other 1/1

3 derivatives checked