ResearchTasksLatency profilingFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedLlama-3.1-8BUniQL7.3TPOT5Feb 26, 2026Qwen 7B Nano 8G 2.5 (inference)UniQL57.7TPOT (ms)3Feb 26, 2026Nemotron-H-8BUniQL6.8TPOT3Feb 26, 2026Mamba2-8B on Nano 8G (inference)UniQL55.3TPOT (Time Per Output Token)2Feb 26, 2026