monroewilliams/Qwen3.8-Flash-Next-oQ4e-fp16-mtp

Same weights, converted vs Jundot/Qwen3.8-Flash-Next-oQ4e-mtp. Same weights stored in another dtype (U32/BF16 → U32/F16).

Tensors changed: lm_head 0/3, Embeddings 0/91, Norms 0/302, other 0/22, Attention 0/1392, MLP 0/1190

On the Hub · Downloads on Model Pulse