Hacker News: flifensteinhttps://news.ycombinator.com/user?id=flifensteinHacker News RSShttps://hnrss.org/hnrss v2.1.1Fri, 31 Jul 2026 01:37:39 +0000<![CDATA[New comment by flifenstein in "Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution"]]>Author here(I'm on the team). We updated this post after Monday's Kimi K3 weights release: fitting the 2.8T model means going from 8×B200 to 8×B300, ~20% more hardware cost, and concurrency drops from 24 to 16 users vs GLM-5.2. Caveat we're upfront about in the post: our 64-task SWEBench Pro subset may be in Kimi's training set, so the 86% resolve rate is an upper bound.

Let us know your thoughts, we really value feedback!

]]>Wed, 29 Jul 2026 14:40:02 +0000https://news.ycombinator.com/item?id=49098144flifensteinhttps://news.ycombinator.com/item?id=49098144https://news.ycombinator.com/item?id=49098144<![CDATA[Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution]]>Article URL: https://aistack.imec-int.com/blog/gpu-self-hosting

Comments URL: https://news.ycombinator.com/item?id=49098130

Points: 151

# Comments: 47

]]>Wed, 29 Jul 2026 14:38:35 +0000https://aistack.imec-int.com/blog/gpu-self-hostingflifensteinhttps://news.ycombinator.com/item?id=49098130https://news.ycombinator.com/item?id=49098130<![CDATA[How many devs can you fit on a GPU?]]>Article URL: https://aistack.imec-int.com/blog/gpu-self-hosting

Comments URL: https://news.ycombinator.com/item?id=49021174

Points: 1

# Comments: 0

]]>Thu, 23 Jul 2026 13:22:22 +0000https://aistack.imec-int.com/blog/gpu-self-hostingflifensteinhttps://news.ycombinator.com/item?id=49021174https://news.ycombinator.com/item?id=49021174