{"id":"2089002532878307738","url":"https://x.com/soyaakinohara5/status/2089002532878307738","text":"https://huggingface.co/soyaakinohara/qwen3.8-27b-abliterated-3.69pbw-12GB-MTP.gguf\nqwen3.8 27bの無検閲版で、日本語の精度をいい感じに残したまま11.7GBまで量子化したモデルを公開しました！！VRAM16GBカード単騎で動作するレベルの27Bとしてはかなり軽量高速で日本語がまともかと思います\n5060Ti 16gb & 3070 8gbの環境で最高39tok/s","author":{"name":"秋ノ原窓也@第五拠点","username":"soyaakinohara5","avatarUrl":"https://pbs.twimg.com/profile_images/2079105510742298624/eAbf_2dg_200x200.jpg"},"createdAt":"Sun Aug 16 14:53:10 +0000 2026","engagement":{"replies":9,"retweets":344,"likes":3171,"views":306906},"externalLink":{"url":"https://huggingface.co/soyaakinohara/qwen3.8-27b-abliterated-3.69pbw-12GB-MTP.gguf","displayUrl":"huggingface.co","title":"soyaakinohara/qwen3.8-27b-abliterated-3.69bpw-12GB-MTP.gguf · Hugging Face","description":"We’re on a journey to advance and democratize artificial intelligence through open source and open science.","thumbnailUrl":"https://cdn-thumbnails.huggingface.co/social-thumbnails/models/soyaakinohara/qwen3.8-27b-abliterated-3.69bpw-12GB-MTP.gguf.png"}}