Qwen3.8-Flash-Next-Abliterated-LoRA cuts refusals on harmful prompts, authors say
AI-generated summary The authors of Qwen3.8-Flash-Next-Abliterated-LoRA announced the release on November 15, 2026, stating that the file size is 2.3 MB and has a rank of rank-1. The authors report that the LoRA reduces refusals on harmful prompts compared to the base Q5 model.