second-state/Llama-4-Scout-17B-16E-Instruct-GGUF

Same weights, converted vs unsloth/Llama-4-Scout-17B-16E-Instruct. Same weights stored in another dtype (BF16 → F16/F32), plus 145 new tensors.

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

On the Hub · Downloads on Model Pulse