mlx-community/gemma-3-12b-it-bf16

Full fine-tune vs google/gemma-3-12b-pt. Fine-tune: 97% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/3, Norms 290/400, MLP 144/252, Attention 192/408, other 1/1

1 derivatives checked

• mlx-community/gemma-3-text-12b-it-4bit: Quantization

On the Hub · Downloads on Model Pulse