ResearchDatasetsHigh-load generation workloadHigh-load generation workloadFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsLLM ServingHigh-load generation workload w/ NVLink443TTFT (ms)27LLM ServingHigh-load generation workload w/o NVLink878Time To First Token (ms)27