ResearchTasksTime To First Token (TTFT) LatencyFollowBenchmarksDataset NameSOTA MethodSortMost resultsRecently updatedMost papersApplyDataset NameSOTA methodMetricTrendResultsLast UpdatedControlled synthetic corpus Config B 1.0 (test)APC+Optimized40.1TTFT p504Jun 19, 2026Controlled synthetic corpus Config A 1.0 (test)APC+Optimized57.9TTFT p50 Latency4Jun 19, 2026Llama 8B 8K Shared Context 3.1Q4 cache reload24.8TTFT (ms)4May 6, 2026Llama 8B 4K Shared Context 3.1Q4 cache reload24.7TTFT (ms)4May 6, 2026Llama 8B 1K Shared Context 3.1FP16 cache copy21.7TTFT (ms)4May 6, 2026