LLM — AI news2 storiesQwen3.8 delivers strong performance on 12GB VRAM r/LocalLLaMAarXiv restricts irresponsible use of LLMs AI 総合（日本語）TopicsArchive