News for hackers. Stuff that matters.

A comment on I think Anthropic and OpenAI have found product-market fit. Parent comment

Re:I think Anthropic and OpenAI have found product-market fit by Tepix on Wednesday May 27, 2026 @11:41PM (#48304214)
Kimi K2.6 does not run well on 256GB.
[ Parent | on HN ]
Re:I think Anthropic and OpenAI have found product-market fit by girvo on Thursday May 28, 2026 @03:23AM (#48305750)
True, I might be thinking of some of the communities four-Spark clusters for it; it’s already int4 right?
[ Parent | on HN ]
Re:I think Anthropic and OpenAI have found product-market fit by Tepix on Saturday May 30, 2026 @03:43AM (#48333721)
Yeah, the default quants are 595GB. Even four Sparks would require a quant lower than 4bit
[ Parent | on HN ]
Re:I think Anthropic and OpenAI have found product-market fit by zozbot234 on Thursday May 28, 2026 @05:23AM (#48306593)
Have you tried it? It would be slow for sure, but the main limitation AIUI would actually be storing the context in RAM - models like Kimi and GLM have high demands there which limit your ability to get meaningful aggregate throughput via large batches.
[ Parent | on HN ]
Re:I think Anthropic and OpenAI have found product-market fit by Tepix on Saturday May 30, 2026 @03:39AM (#48333696)
No need to try really. 1100b weights with 256GB RAM that‘s less than 1.8 bits per weight if you want a little bit of context.

How is that supposed to give good results?

[ Parent | on HN ]