mlx-community/Qwen3.8-27B-bf16
Partial fine-tune vs Qwen/Qwen3.8-27B. Partial change: 0% of the weights changed (norms), in 64 of 64 layers.
Tensors changed: Norms 161/209, Attention 0/400, MLP 0/192, Embeddings 0/1
2 derivatives checked
- bluehawana/Qwen3.8-27B-bf16-MTP-MLX: Same weights, repacked
- faelfernandes/Qwen3.8-27B-3.5bit: Quantization
6 other repos hold byte-for-byte identical weight files.