List your company

Relative Pos. Attention on Llama2-7b (q=32, k=32) (1k)

114.85TFLOPS (Relative Pos. Attention)FlexAttention
10.69437.734564.77591.8155May 6, 2026

Evaluation results

MethodLinks
FlexAttentionImplementation=PyTorch FlexAttention2026.05114.85
CuBridgeImplementation=CuBridge2026.05105.8
Qimeng AttnImplementation=Qimeng Attention2026.0571.05
TorchImplementation=Standard PyTorch2026.0514.7