deepseek-ai

deepseek-ai/flashmla

steady

FlashMLA: Efficient Multi-head Latent Attention Kernels

C++ View on GitHub

Stars

12,953

Forks

1,161

Open issues

75

24h

+7

+0.1%

7d

+18

+0.1%

Refresh

2h

Star history (7 days)

Last checked

1h ago

Last pushed

15 Sep 2026

Next check

just now