List your company

Combo Attention (PrefixLM+Softcap+Sigmoid) on Llama2-7b (q=32, k=32) (1k)

80.03Throughput (TFLOPS)CuBridge
5.46224.82144.1863.539May 6, 2026

Evaluation results

MethodLinks
CuBridgeImplementation=CuBridge2026.0580.03
Qimeng AttnImplementation=Qimeng Attention2026.0517.78
TorchImplementation=Standard PyTorch2026.058.33