digi-texx/Qwen3.8-27B-FULL-NVFP4
Same weights, repacked vs QUASAR-QAT/Qwen3.8-27B-QUASAR-NVFP4. Same weights in different files (re-saved, re-sharded or new metadata), plus 4 new tensors.
Tensors changed: Attention 0/1472, MLP 0/883, Embeddings 0/5, Norms 0/325, other 0/1