Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Masked Multi-Head Attention on T4 GPU Synthetic Performance Benchmark
Loading...
19.07
Performance (TFLOPS)
DeepSeek-V3 + Ours
1.5876
6.1263
10.665
15.2037
Jun 14, 2025
Performance (TFLOPS)
Updated 4d ago
Evaluation Results
Method
Method
Links
Performance (TFLOPS)
DeepSeek-V3 + Ours
Head Dimension=128, Se...
2025.06
19.07
Flex Attention
Head Dimension=64, Seq...
2025.06
13.45
flash-attn v1
Head Dimension=64, Seq...
2025.06
12.81
cuDNN
Head Dimension=64, Seq...
2025.06
8.11
DeepSeek-V3
Head Dimension=128, Se...
2025.06
2.26
Feedback
Search any
task
Search any
task