{
  "model": "Qwen3.8-2.4T-A95B-FP8",
  "engine": "SGLang",
  "hardware": "4 nodes x 8 NVIDIA H20-3e (32 GPUs)",
  "parallelism": "TP8 x PP4",
  "result_scope": {
    "planned_rounds": 62,
    "passed_rounds": 45,
    "failed_rounds": 2,
    "not_executed_rounds": 10,
    "context_limited_rows": 5,
    "completed_formal_requests": 2672,
    "vllm": "NOT_EXECUTED_BY_SCOPE"
  },
  "cases": [
    {
      "case_id": "smoke-128-128-c1",
      "stage": "smoke",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 1,
      "rounds": 1,
      "completed_requests": 4,
      "metrics": {
        "request_throughput": {
          "median": 0.4002428038200527,
          "min": 0.4002428038200527,
          "max": 0.4002428038200527
        },
        "output_throughput": {
          "median": 51.23107888896674,
          "min": 51.23107888896674,
          "max": 51.23107888896674
        },
        "total_token_throughput": {
          "median": 102.46215777793348,
          "min": 102.46215777793348,
          "max": 102.46215777793348
        },
        "mean_ttft_ms": {
          "median": 126.11863890197128,
          "min": 126.11863890197128,
          "max": 126.11863890197128
        },
        "p95_ttft_ms": {
          "median": 129.36211084015667,
          "min": 129.36211084015667,
          "max": 129.36211084015667
        },
        "mean_tpot_ms": {
          "median": 18.67640778829351,
          "min": 18.67640778829351,
          "max": 18.67640778829351
        },
        "p95_tpot_ms": {
          "median": 18.729045326770173,
          "min": 18.729045326770173,
          "max": 18.729045326770173
        },
        "mean_e2el_ms": {
          "median": 2498.022428015247,
          "min": 2498.022428015247,
          "max": 2498.022428015247
        },
        "p95_e2el_ms": {
          "median": 2503.238424891606,
          "min": 2503.238424891606,
          "max": 2503.238424891606
        }
      }
    },
    {
      "case_id": "short-128-128-c1",
      "stage": "baseline",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 1,
      "rounds": 3,
      "completed_requests": 48,
      "metrics": {
        "request_throughput": {
          "median": 0.3984351966257339,
          "min": 0.39803276983192737,
          "max": 0.39919612225171736
        },
        "output_throughput": {
          "median": 50.99970516809394,
          "min": 50.9481945384867,
          "max": 51.09710364821982
        },
        "total_token_throughput": {
          "median": 101.99941033618788,
          "min": 101.8963890769734,
          "max": 102.19420729643964
        },
        "mean_ttft_ms": {
          "median": 127.93344614328817,
          "min": 125.71785887121223,
          "max": 131.10325817251578
        },
        "p95_ttft_ms": {
          "median": 135.59560070279986,
          "min": 133.43765470199287,
          "max": 145.13263606932014
        },
        "mean_tpot_ms": {
          "median": 18.747851271805626,
          "min": 18.714830742949093,
          "max": 18.769988114349765
        },
        "p95_tpot_ms": {
          "median": 18.90874363749662,
          "min": 18.85011715825471,
          "max": 19.099707695963524
        },
        "mean_e2el_ms": {
          "median": 2509.5063493936323,
          "min": 2504.716950497823,
          "max": 2512.08036969183
        },
        "p95_e2el_ms": {
          "median": 2526.5817230101675,
          "min": 2525.3428302239627,
          "max": 2553.2980000134557
        }
      }
    },
    {
      "case_id": "short-128-128-c4",
      "stage": "baseline",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 4,
      "rounds": 3,
      "completed_requests": 192,
      "metrics": {
        "request_throughput": {
          "median": 1.0927290126289966,
          "min": 1.0924531472226517,
          "max": 1.0957373486271587
        },
        "output_throughput": {
          "median": 139.86931361651156,
          "min": 139.83400284449942,
          "max": 140.2543806242763
        },
        "total_token_throughput": {
          "median": 279.7386272330231,
          "min": 279.66800568899885,
          "max": 280.5087612485526
        },
        "mean_ttft_ms": {
          "median": 204.03811974392738,
          "min": 203.91878840746358,
          "max": 204.9189580284292
        },
        "p95_ttft_ms": {
          "median": 234.44278093520552,
          "min": 233.0437646014616,
          "max": 237.08003419451416
        },
        "mean_tpot_ms": {
          "median": 27.209645290845295,
          "min": 27.13379578961208,
          "max": 27.21321742819424
        },
        "p95_tpot_ms": {
          "median": 28.17887856815685,
          "min": 28.106502012173667,
          "max": 28.332863365644663
        },
        "mean_e2el_ms": {
          "median": 3659.543740344816,
          "min": 3650.0301850246615,
          "max": 3660.997571409098
        },
        "p95_e2el_ms": {
          "median": 3794.3418368929997,
          "min": 3733.6750676156953,
          "max": 3822.166770882904
        }
      }
    },
    {
      "case_id": "short-128-128-c8",
      "stage": "baseline",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 8,
      "rounds": 3,
      "completed_requests": 384,
      "metrics": {
        "request_throughput": {
          "median": 1.4873239156086964,
          "min": 1.4857140382105536,
          "max": 1.4877859474029609
        },
        "output_throughput": {
          "median": 190.37746119791314,
          "min": 190.17139689095086,
          "max": 190.436601267579
        },
        "total_token_throughput": {
          "median": 380.75492239582627,
          "min": 380.3427937819017,
          "max": 380.873202535158
        },
        "mean_ttft_ms": {
          "median": 317.1666552007082,
          "min": 315.56042576994514,
          "max": 319.35115818851045
        },
        "p95_ttft_ms": {
          "median": 346.81008071638644,
          "min": 342.41468894761056,
          "max": 356.2275408534333
        },
        "mean_tpot_ms": {
          "median": 39.86125480683611,
          "min": 39.835843152906726,
          "max": 39.876860888384044
        },
        "p95_tpot_ms": {
          "median": 41.236954843112095,
          "min": 41.19269404684992,
          "max": 41.28053573172862
        },
        "mean_e2el_ms": {
          "median": 5377.93978623813,
          "min": 5376.318735619861,
          "max": 5383.712491013284
        },
        "p95_e2el_ms": {
          "median": 5498.1655499199405,
          "min": 5481.451496109366,
          "max": 5502.938569197431
        }
      }
    },
    {
      "case_id": "short-128-128-c16",
      "stage": "baseline",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 16,
      "rounds": 3,
      "completed_requests": 768,
      "metrics": {
        "request_throughput": {
          "median": 2.3069797880447394,
          "min": 2.2473980947215577,
          "max": 2.729928745052605
        },
        "output_throughput": {
          "median": 295.29341286972664,
          "min": 287.6669561243594,
          "max": 349.43087936673345
        },
        "total_token_throughput": {
          "median": 590.5868257394533,
          "min": 575.3339122487188,
          "max": 698.8617587334669
        },
        "mean_ttft_ms": {
          "median": 401.2948538056662,
          "min": 398.84001991231344,
          "max": 414.0759523870656
        },
        "p95_ttft_ms": {
          "median": 558.6350107332692,
          "min": 554.1961338603869,
          "max": 636.5966111188754
        },
        "mean_tpot_ms": {
          "median": 51.457796852639696,
          "min": 42.97856118201586,
          "max": 52.78309649947627
        },
        "p95_tpot_ms": {
          "median": 56.25410009116872,
          "min": 56.18314081931296,
          "max": 57.749854886581694
        },
        "mean_e2el_ms": {
          "median": 6933.980220197554,
          "min": 5859.57212392168,
          "max": 7117.529207820553
        },
        "p95_e2el_ms": {
          "median": 7698.442073422484,
          "min": 7691.097128321417,
          "max": 7891.978862113319
        }
      }
    },
    {
      "case_id": "agent-8k-1k-c1",
      "stage": "baseline",
      "input_tokens": 8192,
      "output_tokens": 1024,
      "concurrency": 1,
      "rounds": 3,
      "completed_requests": 12,
      "metrics": {
        "request_throughput": {
          "median": 0.04749588285226693,
          "min": 0.047488536143804286,
          "max": 0.04754272528701917
        },
        "output_throughput": {
          "median": 48.63578404072133,
          "min": 48.62826101125559,
          "max": 48.68375069390763
        },
        "total_token_throughput": {
          "median": 437.722056366492,
          "min": 437.65434910130034,
          "max": 438.15375624516867
        },
        "mean_ttft_ms": {
          "median": 1636.432361905463,
          "min": 1633.10582539998,
          "max": 1638.4412691695616
        },
        "p95_ttft_ms": {
          "median": 1640.1194979902357,
          "min": 1640.0904579553753,
          "max": 1641.082139289938
        },
        "mean_tpot_ms": {
          "median": 18.979103251871084,
          "min": 18.960684163037065,
          "max": 18.9874380396111
        },
        "p95_tpot_ms": {
          "median": 19.00818573360132,
          "min": 18.992814344567005,
          "max": 19.039166248235066
        },
        "mean_e2el_ms": {
          "median": 21054.06389583368,
          "min": 21033.21226069238,
          "max": 21057.25493992213
        },
        "p95_e2el_ms": {
          "median": 21085.486212116666,
          "min": 21068.074603169225,
          "max": 21109.904268570244
        }
      }
    },
    {
      "case_id": "agent-8k-1k-c4",
      "stage": "baseline",
      "input_tokens": 8192,
      "output_tokens": 1024,
      "concurrency": 4,
      "rounds": 3,
      "completed_requests": 48,
      "metrics": {
        "request_throughput": {
          "median": 0.131702678381208,
          "min": 0.13112323589672825,
          "max": 0.13202419183666383
        },
        "output_throughput": {
          "median": 134.863542662357,
          "min": 134.27019355824973,
          "max": 135.19277244074377
        },
        "total_token_throughput": {
          "median": 1213.771883961213,
          "min": 1208.4317420242478,
          "max": 1216.7349519666939
        },
        "mean_ttft_ms": {
          "median": 2220.3545326774474,
          "min": 2220.225644094171,
          "max": 2222.4691411538515
        },
        "p95_ttft_ms": {
          "median": 2819.1345161758363,
          "min": 2817.9715016158298,
          "max": 2822.1399227622896
        },
        "mean_tpot_ms": {
          "median": 27.515200304054314,
          "min": 27.443048434100096,
          "max": 27.643848944513195
        },
        "p95_tpot_ms": {
          "median": 28.160855076994967,
          "min": 28.071069884417128,
          "max": 28.328626582085835
        },
        "mean_e2el_ms": {
          "median": 30368.404443725012,
          "min": 30294.464192178566,
          "max": 30502.12661139085
        },
        "p95_e2el_ms": {
          "median": 30700.777022517286,
          "min": 30440.107466769405,
          "max": 30702.248659916222
        }
      }
    },
    {
      "case_id": "agent-8k-1k-c8",
      "stage": "baseline",
      "input_tokens": 8192,
      "output_tokens": 1024,
      "concurrency": 8,
      "rounds": 3,
      "completed_requests": 96,
      "metrics": {
        "request_throughput": {
          "median": 0.2260717023243467,
          "min": 0.22463453276231451,
          "max": 0.2266022068974271
        },
        "output_throughput": {
          "median": 231.49742318013102,
          "min": 230.02576154861006,
          "max": 232.04065986296536
        },
        "total_token_throughput": {
          "median": 2083.4768086211793,
          "min": 2070.2318539374905,
          "max": 2088.365938766688
        },
        "mean_ttft_ms": {
          "median": 3011.0449328058166,
          "min": 3010.5397277802695,
          "max": 3013.453338324325
        },
        "p95_ttft_ms": {
          "median": 4390.008029504679,
          "min": 4383.700305921957,
          "max": 4390.623911703005
        },
        "mean_tpot_ms": {
          "median": 31.641996890189372,
          "min": 31.56324981654962,
          "max": 31.866006117753585
        },
        "p95_tpot_ms": {
          "median": 33.16794376750226,
          "min": 32.97583688080552,
          "max": 33.23578939186855
        },
        "mean_e2el_ms": {
          "median": 35383.216156988055,
          "min": 35300.249495136086,
          "max": 35609.46398624219
        },
        "p95_e2el_ms": {
          "median": 35759.98356547207,
          "min": 35758.434898010455,
          "max": 35805.114023783244
        }
      }
    },
    {
      "case_id": "rag-4k-256-c4",
      "stage": "baseline",
      "input_tokens": 4096,
      "output_tokens": 256,
      "concurrency": 4,
      "rounds": 3,
      "completed_requests": 192,
      "metrics": {
        "request_throughput": {
          "median": 0.4960255530167065,
          "min": 0.4953025605406269,
          "max": 0.4969114172286591
        },
        "output_throughput": {
          "median": 126.98254157227686,
          "min": 126.79745549840048,
          "max": 127.20932281053673
        },
        "total_token_throughput": {
          "median": 2158.7032067287064,
          "min": 2155.556743472808,
          "max": 2162.5584877791243
        },
        "mean_ttft_ms": {
          "median": 1665.9018880673102,
          "min": 1665.279541957716,
          "max": 1670.3049206189462
        },
        "p95_ttft_ms": {
          "median": 2015.0486186845228,
          "min": 2014.9696272565052,
          "max": 2015.0710275862366
        },
        "mean_tpot_ms": {
          "median": 25.088109124786282,
          "min": 25.034295412906964,
          "max": 25.11682568886089
        },
        "p95_tpot_ms": {
          "median": 27.80369134428089,
          "min": 27.732093942479466,
          "max": 27.920550207570926
        },
        "mean_e2el_ms": {
          "median": 8063.369714887813,
          "min": 8049.024872248992,
          "max": 8075.095471278473
        },
        "p95_e2el_ms": {
          "median": 8293.217238457872,
          "min": 8220.64230851829,
          "max": 8298.03804194089
        }
      }
    },
    {
      "case_id": "rag-4k-256-c8",
      "stage": "baseline",
      "input_tokens": 4096,
      "output_tokens": 256,
      "concurrency": 8,
      "rounds": 1,
      "completed_requests": 128,
      "metrics": {
        "request_throughput": {
          "median": 0.7662176129293546,
          "min": 0.7662176129293546,
          "max": 0.7662176129293546
        },
        "output_throughput": {
          "median": 196.1517089099148,
          "min": 196.1517089099148,
          "max": 196.1517089099148
        },
        "total_token_throughput": {
          "median": 3334.5790514685514,
          "min": 3334.5790514685514,
          "max": 3334.5790514685514
        },
        "mean_ttft_ms": {
          "median": 2177.9094463017827,
          "min": 2177.9094463017827,
          "max": 2177.9094463017827
        },
        "p95_ttft_ms": {
          "median": 2790.6565761193633,
          "min": 2790.6565761193633,
          "max": 2790.6565761193633
        },
        "mean_tpot_ms": {
          "median": 32.399093890197435,
          "min": 32.399093890197435,
          "max": 32.399093890197435
        },
        "p95_tpot_ms": {
          "median": 35.79091523554834,
          "min": 35.79091523554834,
          "max": 35.79091523554834
        },
        "mean_e2el_ms": {
          "median": 10439.678388302127,
          "min": 10439.678388302127,
          "max": 10439.678388302127
        },
        "p95_e2el_ms": {
          "median": 10773.952074511908,
          "min": 10773.952074511908,
          "max": 10773.952074511908
        }
      }
    },
    {
      "case_id": "rag-16k-256-c1",
      "stage": "prefill",
      "input_tokens": 16384,
      "output_tokens": 256,
      "concurrency": 1,
      "rounds": 3,
      "completed_requests": 12,
      "metrics": {
        "request_throughput": {
          "median": 0.14157220134061071,
          "min": 0.14148922905311673,
          "max": 0.1417798299884528
        },
        "output_throughput": {
          "median": 36.24248354319634,
          "min": 36.22124263759788,
          "max": 36.295636477043914
        },
        "total_token_throughput": {
          "median": 2355.761430307762,
          "min": 2354.380771443862,
          "max": 2359.2163710078544
        },
        "mean_ttft_ms": {
          "median": 2158.647161326371,
          "min": 2155.1552097080275,
          "max": 2164.7006592247635
        },
        "p95_ttft_ms": {
          "median": 2166.6261999635026,
          "min": 2165.312159433961,
          "max": 2170.37127611693
        },
        "mean_tpot_ms": {
          "median": 19.20964052060656,
          "min": 19.192402067995978,
          "max": 19.26323095583083
        },
        "p95_tpot_ms": {
          "median": 19.24021636866325,
          "min": 19.233708637675235,
          "max": 19.324073409296428
        },
        "mean_e2el_ms": {
          "median": 7063.158991979435,
          "min": 7052.709688665345,
          "max": 7067.279103444889
        },
        "p95_e2el_ms": {
          "median": 7069.615903077647,
          "min": 7069.412807631306,
          "max": 7077.2470311960205
        }
      }
    },
    {
      "case_id": "rag-16k-256-c4",
      "stage": "prefill",
      "input_tokens": 16384,
      "output_tokens": 256,
      "concurrency": 4,
      "rounds": 3,
      "completed_requests": 48,
      "metrics": {
        "request_throughput": {
          "median": 0.3846038872704328,
          "min": 0.37975843655702274,
          "max": 0.38790712775103703
        },
        "output_throughput": {
          "median": 98.4585951412308,
          "min": 97.21815975859782,
          "max": 99.30422470426548
        },
        "total_token_throughput": {
          "median": 6399.808684180002,
          "min": 6319.180384308859,
          "max": 6454.774605777256
        },
        "mean_ttft_ms": {
          "median": 3389.821106567979,
          "min": 3386.8259506416507,
          "max": 3394.2863919364754
        },
        "p95_ttft_ms": {
          "median": 4629.219308611937,
          "min": 4623.69808962103,
          "max": 4632.435084669851
        },
        "mean_tpot_ms": {
          "median": 27.48620192938046,
          "min": 27.150428031945562,
          "max": 27.98495425285735
        },
        "p95_tpot_ms": {
          "median": 32.40553471624997,
          "min": 32.36162290022727,
          "max": 33.00971677024648
        },
        "mean_e2el_ms": {
          "median": 10398.802598559996,
          "min": 10310.18509878777,
          "max": 10530.4497264151
        },
        "p95_e2el_ms": {
          "median": 10561.8656007573,
          "min": 10530.095011228696,
          "max": 10683.558649732731
        }
      }
    },
    {
      "case_id": "rag-16k-256-c8",
      "stage": "prefill",
      "input_tokens": 16384,
      "output_tokens": 256,
      "concurrency": 8,
      "rounds": 3,
      "completed_requests": 96,
      "metrics": {
        "request_throughput": {
          "median": 0.5298238437431914,
          "min": 0.5261398205499086,
          "max": 0.5303992692136629
        },
        "output_throughput": {
          "median": 135.634903998257,
          "min": 134.6917940607766,
          "max": 135.7822129186977
        },
        "total_token_throughput": {
          "median": 8816.268759886703,
          "min": 8754.966613950479,
          "max": 8825.84383971535
        },
        "mean_ttft_ms": {
          "median": 5103.144982203958,
          "min": 5066.414021057426,
          "max": 5207.9617764102295
        },
        "p95_ttft_ms": {
          "median": 8092.364924983121,
          "min": 7960.687731276266,
          "max": 8102.271855971775
        },
        "mean_tpot_ms": {
          "median": 39.27129828336868,
          "min": 38.780449609918165,
          "max": 39.60084877423166
        },
        "p95_tpot_ms": {
          "median": 50.70723849942214,
          "min": 50.66365067189669,
          "max": 51.056802802006985
        },
        "mean_e2el_ms": {
          "median": 15096.976426939364,
          "min": 15080.595083316439,
          "max": 15201.36141963303
        },
        "p95_e2el_ms": {
          "median": 15306.243107677437,
          "min": 15297.9935585754,
          "max": 15364.028520649299
        }
      }
    },
    {
      "case_id": "decode-128-2k-c1",
      "stage": "decode",
      "input_tokens": 128,
      "output_tokens": 2048,
      "concurrency": 1,
      "rounds": 3,
      "completed_requests": 12,
      "metrics": {
        "request_throughput": {
          "median": 0.025910014623559095,
          "min": 0.025856952445737616,
          "max": 0.025915150215270813
        },
        "output_throughput": {
          "median": 53.06370994904903,
          "min": 52.95503860887064,
          "max": 53.074227640874625
        },
        "total_token_throughput": {
          "median": 56.38019182086459,
          "min": 56.26472852192506,
          "max": 56.39136686842929
        },
        "mean_ttft_ms": {
          "median": 125.37228572182357,
          "min": 124.00193023495376,
          "max": 126.59330072347075
        },
        "p95_ttft_ms": {
          "median": 126.92641350440681,
          "min": 126.08962573576719,
          "max": 129.35111590195447
        },
        "mean_tpot_ms": {
          "median": 18.79299234775858,
          "min": 18.78993829211123,
          "max": 18.831108434753183
        },
        "p95_tpot_ms": {
          "median": 18.85750265616849,
          "min": 18.832427209323257,
          "max": 18.908644326688187
        },
        "mean_e2el_ms": {
          "median": 38594.62762158364,
          "min": 38587.00561418664,
          "max": 38673.87226666324
        },
        "p95_e2el_ms": {
          "median": 38725.34036063589,
          "min": 38676.068123220466,
          "max": 38829.77888183668
        }
      }
    },
    {
      "case_id": "decode-128-2k-c4",
      "stage": "decode",
      "input_tokens": 128,
      "output_tokens": 2048,
      "concurrency": 4,
      "rounds": 3,
      "completed_requests": 48,
      "metrics": {
        "request_throughput": {
          "median": 0.0712914150180247,
          "min": 0.07107391839621295,
          "max": 0.0719189890107167
        },
        "output_throughput": {
          "median": 146.0048179569146,
          "min": 145.55938487544412,
          "max": 147.2900894939478
        },
        "total_token_throughput": {
          "median": 155.13011907922177,
          "min": 154.6568464301594,
          "max": 156.4957200873195
        },
        "mean_ttft_ms": {
          "median": 205.32086686580442,
          "min": 196.57257097424008,
          "max": 210.13968155602925
        },
        "p95_ttft_ms": {
          "median": 236.72201961744577,
          "min": 219.916699687019,
          "max": 249.28623449523002
        },
        "mean_tpot_ms": {
          "median": 27.306382337242272,
          "min": 27.073853020886126,
          "max": 27.392594541917692
        },
        "p95_tpot_ms": {
          "median": 28.299700700642443,
          "min": 27.886006622732186,
          "max": 28.473227979530254
        },
        "mean_e2el_ms": {
          "median": 56106.30432589096,
          "min": 55616.749704728136,
          "max": 56277.96189417131
        },
        "p95_e2el_ms": {
          "median": 58129.289024393074,
          "min": 57260.896133608185,
          "max": 58498.66893352009
        }
      }
    },
    {
      "case_id": "decode-128-2k-c8",
      "stage": "decode",
      "input_tokens": 128,
      "output_tokens": 2048,
      "concurrency": 8,
      "rounds": 3,
      "completed_requests": 72,
      "metrics": {
        "request_throughput": {
          "median": 0.09811628524720671,
          "min": 0.0955094389857361,
          "max": 0.09883459579646413
        },
        "output_throughput": {
          "median": 200.94215218627934,
          "min": 195.60333104278754,
          "max": 202.41325219115853
        },
        "total_token_throughput": {
          "median": 213.50103669792182,
          "min": 207.82853923296176,
          "max": 215.06408045310593
        },
        "mean_ttft_ms": {
          "median": 314.898140670266,
          "min": 313.0695012902531,
          "max": 316.4222938939929
        },
        "p95_ttft_ms": {
          "median": 341.451816027984,
          "min": 338.3156476309523,
          "max": 346.4752336265519
        },
        "mean_tpot_ms": {
          "median": 39.677700033311524,
          "min": 39.38655481269271,
          "max": 40.7639247831033
        },
        "p95_tpot_ms": {
          "median": 40.18494039131296,
          "min": 39.728716248111795,
          "max": 41.03959487518181
        },
        "mean_e2el_ms": {
          "median": 81533.32146947893,
          "min": 80940.69999547598,
          "max": 83758.65217168273
        },
        "p95_e2el_ms": {
          "median": 82458.30807539169,
          "min": 81664.34307249729,
          "max": 84337.38673434127
        }
      }
    },
    {
      "case_id": "short-128-128-c32",
      "stage": "saturation",
      "input_tokens": 128,
      "output_tokens": 128,
      "concurrency": 32,
      "rounds": 1,
      "completed_requests": 512,
      "metrics": {
        "request_throughput": {
          "median": 3.846731241266357,
          "min": 3.846731241266357,
          "max": 3.846731241266357
        },
        "output_throughput": {
          "median": 492.38159888209367,
          "min": 492.38159888209367,
          "max": 492.38159888209367
        },
        "total_token_throughput": {
          "median": 984.7631977641873,
          "min": 984.7631977641873,
          "max": 984.7631977641873
        },
        "mean_ttft_ms": {
          "median": 520.5021024139569,
          "min": 520.5021024139569,
          "max": 520.5021024139569
        },
        "p95_ttft_ms": {
          "median": 701.9454820081592,
          "min": 701.9454820081592,
          "max": 701.9454820081592
        },
        "mean_tpot_ms": {
          "median": 61.277402446605954,
          "min": 61.277402446605954,
          "max": 61.277402446605954
        },
        "p95_tpot_ms": {
          "median": 65.36672783861539,
          "min": 65.36672783861539,
          "max": 65.36672783861539
        },
        "mean_e2el_ms": {
          "median": 8302.732213132913,
          "min": 8302.732213132913,
          "max": 8302.732213132913
        },
        "p95_e2el_ms": {
          "median": 8692.103337403387,
          "min": 8692.103337403387,
          "max": 8692.103337403387
        }
      }
    }
  ],
  "contexts": [
    {
      "configured_context_tokens": 65536,
      "server_prompt_tokens": 63464,
      "completion_tokens": 121,
      "reasoning_tokens": 104,
      "needle_depth": 0.5,
      "status": "PASS",
      "result_sha256": "ce996524d036450daf5d59c0002480cff9f14c0c8969fd57910f0bdc0d917a89"
    },
    {
      "configured_context_tokens": 131072,
      "server_prompt_tokens": 129001,
      "completion_tokens": 95,
      "reasoning_tokens": 77,
      "needle_depth": 0.5,
      "status": "PASS",
      "result_sha256": "07169a501d8e9b34fb01428764a10ed468845e37d539bedab4a775d499e4c064"
    },
    {
      "configured_context_tokens": 262144,
      "server_prompt_tokens": 260073,
      "completion_tokens": 95,
      "reasoning_tokens": 77,
      "needle_depth": 0.5,
      "status": "PASS",
      "result_sha256": "9eade32ffd6b9ec26ba6c53ddd61b00a2fa6eae421f44abd173aac57097a6aac"
    }
  ],
  "telemetry": {
    "window": "full SGLang performance-suite window",
    "gpu": {
      "utilization_percent": {
        "samples": 15392,
        "mean": 54.88351091476091,
        "p50": 60.0,
        "p95": 99.0,
        "max": 100.0
      },
      "framebuffer_used_mib": {
        "samples": 15392,
        "mean": 126837.41294178794,
        "p50": 128853.0,
        "p95": 129395.0,
        "max": 129395.0
      },
      "power_watts": {
        "samples": 15392,
        "mean": 223.42793886434512,
        "p50": 231.049,
        "p95": 297.20259999999985,
        "max": 423.588
      },
      "sm_clock_mhz": {
        "samples": 15392,
        "mean": 1980.0,
        "p50": 1980.0,
        "p95": 1980.0,
        "max": 1980.0
      }
    },
    "client": {
      "sliding_p95_ttft_seconds": {
        "samples": 376,
        "mean": 2.1813470949921814,
        "p50": 0.9741935483870967,
        "p95": 6.024999999999999,
        "max": 9.6
      },
      "sliding_p95_tpot_seconds": {
        "samples": 376,
        "mean": 0.04006391116298411,
        "p50": 0.0295,
        "p95": 0.07825376506024097,
        "max": 0.0785
      },
      "successful_output_tokens_per_second": {
        "samples": 477,
        "mean": 106.51064079210796,
        "p50": 74.47272727272727,
        "p95": 297.8909090909091,
        "max": 521.3090909090909
      },
      "sliding_p95_e2e_seconds": {
        "samples": 376,
        "mean": 36.279265648443186,
        "p50": 28.999999999999996,
        "p95": 117.0,
        "max": 117.0
      }
    },
    "rdma": {
      "transmit_bytes_per_second": {
        "samples": 1924,
        "mean": 0.026875969039215672,
        "p50": 0.008030675453218547,
        "p95": 0.11374041571535846,
        "max": 0.2961927856911312,
        "unit": "GiB/s per sampled node series"
      },
      "receive_bytes_per_second": {
        "samples": 1924,
        "mean": 0.02684996329134033,
        "p50": 0.00803044739771973,
        "p95": 0.11362512755800376,
        "max": 0.2915880594741214,
        "unit": "GiB/s per sampled node series"
      }
    }
  },
  "source_receipts": {
    "suite_status_sha256": "5b4bcf9bfd8bf0d37b884862f483c3e5b7b940c30aac1c268e60bc5508b3bbd2",
    "grafana_evidence_sha256": "a28da1f3a28af1f23b4aee544ce7ec7cfcca2db299fcdccb89632a15be6ce36f"
  }
}
