gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090-LMHead4

Same weights, repacked vs gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090. Same weights in different files (re-saved, re-sharded or new metadata), plus 15 new tensors.

Tensors changed: Attention 0/1180, MLP 0/880, Embeddings 0/4, Norms 0/319, lm_head 0/4

2 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse