sluttybutfast/KAT-Coder-V2.5-Dev-Vision-OptiQ-4bit

Same weights, repacked vs mlx-community/KAT-Coder-V2.5-Dev-OptiQ-4bit. Same weights in different files (re-saved, re-sharded or new metadata), plus 333 new tensors.

Tensors changed: Embeddings 0/3, Norms 0/131, Attention 0/660, MLP 0/960, lm_head 0/3

1 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse