llmfan46/Qwen3.5-27B-uncensored-heretic-v2-Native-MTP-Preserved-NVFP4-GGUF

Same weights, converted vs llmfan46/Qwen3.5-27B-uncensored-heretic-v2-Native-MTP-Preserved. Same weights stored in another dtype (BF16 → T40/BF16), plus 559 new tensors.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/161

On the Hub · Downloads on Model Pulse