VIDraft/Darwin-28B-KOREA

Full fine-tune vs FINAL-Bench/Darwin-28B-Opus. Fine-tune: 98% of the weights changed, across 64 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 2/5, Norms 324/325, Attention 446/560, MLP 195/307, other 1/1

2 derivatives checked

• mradermacher/Darwin-28B-KOREA-i1-GGUF: Quantization
• mradermacher/Darwin-28B-KOREA-GGUF: Quantization

On the Hub · Downloads on Model Pulse