NANI-Nithin/Instella-MoE-16B-A3B-Think-GGUF

Same weights, converted vs amd/Instella-MoE-16B-A3B-Think. Same weights stored in another dtype (BF16/F32 → BF16/F32), plus 185 new tensors.

Tensors changed: Embeddings 0/1, lm_head 0/1, Norms 0/82, MLP 0/107, Attention 0/81

On the Hub · Downloads on Model Pulse