ResearchTasksMasked Multi-Head AttentionMasked Multi-Head AttentionFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedT4 GPU Synthetic Performance BenchmarkDeepSeek-V3 + Ours19.07Performance (TFLOPS)5Feb 26, 2026