lmstudio-community/Qwen3-VL-2B-Instruct-MLX-bf16
Same weights, repacked vs Qwen/Qwen3-VL-2B-Instruct. Same weights in different files (re-saved, re-sharded or new metadata), plus 310 new tensors.
Tensors changed: Attention 0/96, MLP 0/112, Norms 0/104, Embeddings 0/2
1 other repos hold byte-for-byte identical weight files.