Sudo su (@sudoingX)단일 24GB GPU 환경에서는 Qwen 3.6 27B dense를 Q4로 올리는 것이 사실상 최선의 선택이라는 경험담입니다. 3090/4090/7900 XTX 급 GPU에서 충분한 컨텍스트 여유를 확보하면서도, 작은 모델들이 잃기 쉬운 추론력을 더 잘 유지한다고 평가합니다.https://x.com/sudoingX/status/2055679136501501978#qwen #llm #gpu #quantization
Related
“What’s the road block? The promised #productivity gains promised are not being realised.” — Internet Of Bugs #AI / #Gen...
“What’s the road block? The promised #productivity gains promised are not being realised.” — Internet Of Bugs #AI / #GenerativeAI / #bubble / #AIBubble vs #DotComBubble
Samsung’s flagship lineup may grow to 8 phones a year with the Galaxy Z TriFold 2The Galaxy Z TriFold 2 and Galaxy S27 P...
Samsung’s flagship lineup may grow to 8 phones a year with the Galaxy Z TriFold 2The Galaxy Z TriFold 2 and Galaxy S27 Pro could help Samsung build its most diverse flagship smartp...
🤖 Need a thing“I’ve been collaborating with AI on my music, and I just dropped a new track called ‘Need a Thing.’ It’s a...
🤖 Need a thing“I’ve been collaborating with AI on my music, and I just dropped a new track called ‘Need a Thing.’ It’s a rebuttal to Rihanna’s ‘Needed Me’ featuring TWO different A...