ResearchTasksGPU serving throughputFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedControlled traces 8 streams 4096 prompt tokens 512 decode tokensRW-TTT full274.61Throughput (tok/s)6May 28, 2026
Controlled traces 8 streams 4096 prompt tokens 512 decode tokensRW-TTT full274.61Throughput (tok/s)6May 28, 2026