News for hackers. Stuff that matters.

A comment on Qwen3.8-Max: A New Bar for Coding and Cowork. Parent comment

Re:Qwen3.8-Max: A New Bar for Coding and Cowork by colordrops on Monday August 03, 2026 @02:37AM (#49151994)
MoE models can use system memory along with a GPU.
[ Parent | on HN ]
Re:Qwen3.8-Max: A New Bar for Coding and Cowork by formvoltron on Monday August 03, 2026 @11:18AM (#49156936)
and get high token bandwidth?
[ Parent | on HN ]
Re:Qwen3.8-Max: A New Bar for Coding and Cowork by colordrops on Monday August 03, 2026 @02:00PM (#49159294)
Similar to a spark, which isn't blazing fast but usable.
[ Parent | on HN ]
Re:Qwen3.8-Max: A New Bar for Coding and Cowork by numpad0 on Monday August 03, 2026 @03:25PM (#49160292)
Not badly so because MoE models(identifiable by "CoolName-xxxB-AxxB" naming scheme) have bunch of branches in the middle that only one out of all gets non-zero values. Each of branches aka "Experts" as well as top/bottom parts are significantly smaller than the whole, and so CPU emulation of CUDA operations mixed with GPU taking as much as possible become not so out of question, unlike for dense models("CoolName-xxxB" without "-AxxB")
[ Parent | on HN ]