N
Hacker Next
new
past
show
ask
show
jobs
submit
login
▲
MiniMax teased M3 Sparse Attention: 9.7x prefilling, 15.6x decoding at 1M
(
twitter.com
)
8 points by
rebekkamikkoa
18 hours ago
|
0 comments
add comment
Rendered at 15:04:37 GMT+0000 (Coordinated Universal Time) with Vercel.