MK4-Research/MK4-Ginko-v1

Partial fine-tune vs mlx-community/Qwen3.8-27B-4bit. Partial change: 70% of the weights changed (attention, mlp), in 64 of 64 layers.

Tensors changed: Embeddings 0/3, Norms 0/209, Attention 192/1056, MLP 576/576, lm_head 0/3

On the Hub · Downloads on Model Pulse