murillo2000/Kimi-K3-0.40B-GGUF

Same weights, converted vs inference-optimization/Kimi-K3-0.40B. Same weights stored in another dtype (F32 → F16/F32), plus 84 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/27, MLP 0/24, Attention 0/56, other 0/7

On the Hub · Downloads on Model Pulse