News for hackers. Stuff that matters.

A comment on Kimi K3.1 Pro weights out. Parent comment

Re:Kimi K3.1 Pro weights out by vorticalbox on Wednesday November 04, 2026 @11:43AM (#50368183)
Do you need 450k context?

I usually keep under 200k if I can help it for Claude/ open ai

And lower locally on my Mac to save on prefill.

[ Parent | on HN ]
Re:Kimi K3.1 Pro weights out by __jf__ on Wednesday November 04, 2026 @12:53PM (#50369123)
No, also because that would mean I have to read up on rope scaling or whatever to go beyond the default 256k context, which so far I’ve avoided. But you can allocate it as 220k context for 2 parallel Pi workers, that really speeds things up! Or as ~8x 50k context for dedicated parallel document processing.
[ Parent | on HN ]
Re:Kimi K3.1 Pro weights out by octoberfranklin on Thursday November 05, 2026 @10:01AM (#50379952)
Kimi is NoPE not RoPE.
[ Parent | on HN ]