unsloth/cogito-v2-preview-deepseek-671B-MoE-GGUF

Same weights, converted vs deepcogito/cogito-v2-preview-deepseek-671B-MoE. Same weights stored in another dtype (BF16 → BF16/F32), plus 354 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/245, MLP 0/241, Attention 0/244

On the Hub · Downloads on Model Pulse