mlx-community/Qwen3-235B-A22B-Instruct-2507-3bit-DWQ

Partial fine-tune vs mlx-community/Qwen3-235B-A22B-Instruct-2507-8bit. Partial change: 51% of the weights changed (attention, embeddings, lm_head, mlp, norms), in 94 of 94 layers.

Tensors changed: Embeddings 1/2, Norms 148/377, MLP 388/752, Attention 392/752, lm_head 1/2

On the Hub · Downloads on Model Pulse