ResearchDatasetsHigh-load generation workloadFollowBenchmarksTask NameDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyTask NameDataset NameSOTA ResultTrendResultsLLM ServingHigh-load generation workload w/ NVLink443TTFT (ms)27LLM ServingHigh-load generation workload w/o NVLink878Time To First Token (ms)27