Well, Cerebras pretty much needs a data center to simply fit the 405B model for ... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		pama 8 months ago \| parent \| context \| favorite \| on: Llama 3.1 405B now runs at 969 tokens/s on Cerebra... Well, Cerebras pretty much needs a data center to simply fit the 405B model for inference.

throwawaymaths 8 months ago [–]

I guess this just shows the insanity of venture led AI hardware hype and shady startup messaging practices

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact