News for hackers. Stuff that matters.

A comment on DeepSeek V4.1 Flash 1017 (Max). Parent comment

Re:DeepSeek V4.1 Flash 1017 (Max) by swiftcoder on Monday October 19, 2026 @04:22AM (#50149759)
This has to be a distilled model? I don't see any route to 200 tokens/s without serious datacenter hardware
[ Parent | on HN ]
Re:DeepSeek V4.1 Flash 1017 (Max) by naraic0o on Monday October 19, 2026 @05:58AM (#50150422)
you mean quantized? it’s only got 16B active parameters so it can be pretty fast at decode.
[ Parent | on HN ]
Re:DeepSeek V4.1 Flash 1017 (Max) by swiftcoder on Monday October 19, 2026 @06:36AM (#50150687)
sorry, yes, quantised
[ Parent | on HN ]