# Wed Oct 28 02:25:20 PM +03 2026 mode=batch server=127.0.0.1:18020
ROW 64conc 128in/512out | e2e=2376.18 tok/s | decode(C/medTPOT)=2556 | medTPOT=25.04 ms | meanTTFT=971.22 ms | dur=55.16s
ROW 64conc 256in/256out | e2e=1787.20 tok/s | decode(C/medTPOT)=2008 | medTPOT=31.87 ms | meanTTFT=1189.18 ms | dur=36.67s
ROW cohort C1 real prompts T=default | e2e=78.26 tok/s | decode(C/meanTPOT)=78.7 | tok/step=- | meanTTFT=80.77 ms
ROW cohort C2 real prompts T=default | e2e=137.19 tok/s | decode(C/meanTPOT)=143.1 | tok/step=- | meanTTFT=168.48 ms
ROW cohort C4 real prompts T=default | e2e=274.93 tok/s | decode(C/meanTPOT)=280.3 | tok/step=- | meanTTFT=169.65 ms
ROW cohort C8 real prompts T=default | e2e=600.31 tok/s | decode(C/meanTPOT)=642.1 | tok/step=- | meanTTFT=251.18 ms
# raw logs in /home/zeka/ai/benchmarks/hyperqwen-5090-2026-10-28/results/batch/run1
