(评论)
(comments)

原始链接: https://news.ycombinator.com/item?id=43427115

谷歌声称其新的Gemma 3是能在单GPU上运行的最强大的AI模型。Hacker News用户CamperBob2对此提出了质疑,提到了DeepSeek R1的1.58位Unsloth量化,该模型可以在只有128MB系统内存的单GPU上运行,尽管速度较慢。其他用户也加入了讨论,质疑极低比特量化的实用性,澄清了内存容量,并强调了由于AI技术的快速发展,此类说法必然短暂的特性。一些人甚至质疑Gemma 3是否是真正最强大且可以在单GPU上运行的模型。总的来说,讨论集中在创建最有效和高性能AI模型的持续竞争上,重点关注资源限制和不断发展的AI环境。


原文
Hacker News new | past | comments | ask | show | jobs | submit login
Google calls Gemma 3 the most powerful AI model you can run on one GPU (theverge.com)
22 points by gmays 1 hour ago | hide | past | favorite | 6 comments










Technically, the 1.58-bit Unsloth quant of DeepSeek R1 runs on a single GPU+128MB of system RAM. It performs amazingly well, but you'd better not be in a hurry.


Have you tried the 0.000000001 bit quant? IIUC it's a single param so youd get much faster speeds.


You probably meant 128GB


...until this coming tuesday? ...let's talk value.


It’s a 27B model, I highly doubt that.


Which part do you doubt? That it is the most powerful? That it runs on a single GPU?






Join us for AI Startup School this June 16-17 in San Francisco!


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact



Search:
联系我们 contact @ memedata.com