Y
Hacker News
new
|
ask
|
show
|
jobs
MiniMax teased M3 Sparse Attention: 9.7x prefilling, 15.6x decoding at 1M
(
twitter.com
)
9 points
by
rebekkamikkoa
25 days ago