nightmedia/Qwen3.5-27B-Engineer-Deckard-Gemini

Full fine-tune vs Jackrong/Qwen3.5-27B-Gemini-3.1-Pro-Reasoning-Distill. Fine-tune: 99% of the weights changed, across 64 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 4/5, MLP 304/307, Attention 556/560, Norms 319/325

2 derivatives checked