Hacker News new | ask | show | jobs
by zekrioca 284 days ago
Paper: https://arxiv.org/abs/2509.05276v1
1 comments
BoiledCabbage 284 days ago
The paper appears to list the 100x speed-up as time to first token. As I understand that doesn't imply 100x in throughout. Is there more listed in the paper itself?
link