ResearchTasksAttention Operator ThroughputAttention Operator ThroughputFollowBenchmarksDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyDatasetSOTA methodMetricTrendResultsLast updatedLlama 405B (128 Q-heads/8 KV-heads/128 Head-dimension) 3.1CuBridge615.39TFLOPS62May 7, 2026Llama2 7B (32 Q-heads/32 KV-heads/128 Head-dimension)flash-attn v2207.3Attention TFLOPS42May 7, 2026Qwen2.5 72B (64 Q-heads/8 KV-heads/128 Head-dimension)flash-attn v2222.5Attention Throughput (TFLOPS)29May 7, 2026
Qwen2.5 72B (64 Q-heads/8 KV-heads/128 Head-dimension)flash-attn v2222.5Attention Throughput (TFLOPS)29May 7, 2026