unsloth/cogito-v2-preview-llama-109B-MoE-GGUF

Same weights, converted vs deepcogito/cogito-v2-preview-llama-109B-MoE. Same weights stored in another dtype (BF16 → BF16/F32), plus 145 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, MLP 0/192, Norms 0/97, Attention 0/192

On the Hub · Downloads on Model Pulse