third-intelligence/llm-jp-4-kappa-32b-a3b-v0.1

Full fine-tune vs llm-jp/llm-jp-4-32b-a3b-thinking. Fine-tune: 99% of the weights changed, across 32 of 32 layers.

Tensors changed: Embeddings 1/1, Norms 117/129, MLP 2722/2741, Attention 128/128, lm_head 1/1

1 derivatives checked

• mmnga-o/llm-jp-4-kappa-32b-a3b-v0.1-gguf: Quantization

On the Hub · Downloads on Model Pulse