ResearchTasksAgentic ServingFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedShareGPT vLLM native multi-turn benchmark 32 concurrent chat sessionsUltraQuant0.86P50 TTFT (Warm Rounds, r2-3) Ratio1Jun 19, 2026
ShareGPT vLLM native multi-turn benchmark 32 concurrent chat sessionsUltraQuant0.86P50 TTFT (Warm Rounds, r2-3) Ratio1Jun 19, 2026