Every story and comment from the Hacker News API, in a layout borrowed from 2001.
A comment on Accelerating GPT-5.6 Sol Ultrafast. Parent comment
Batching works because of severe memory bottleneck, but Cerebras whole thing is serving models out of "L1 cache" (?).
There are no comments on this comment yet.
Batching works because of severe memory bottleneck, but Cerebras whole thing is serving models out of "L1 cache" (?).