Top New Trends Comments Past Ask Show Jobs

MiniMax teased M3 Sparse Attention: 9.7x prefilling, 15.6x decoding at 1M(twitter.com)

9 points·by rebekkamikkoa·vor 2 Monaten·0 comments

MiniMax teased M3 Sparse Attention: 9.7x prefilling, 15.6x decoding at 1M

https://twitter.com/SkylerMiao7/status/2059285750458544561

0 comments

—

MiniMax teased M3 Sparse Attention: 9.7x prefilling, 15.6x decoding at 1M · HackerTrans