{
  "view": "inference",
  "apiVersion": "v1",
  "params": {
    "model": "DeepSeek-V4-Pro",
    "sequence": "8k/1k",
    "precisions": [
      "fp4"
    ],
    "metric": "y_tpPerGpu",
    "xmode": "interactivity",
    "xmetric": "p90_ttft",
    "percentile": "p90",
    "date": "2026-07-31",
    "runId": null,
    "gpus": [
      "b200",
      "b300"
    ],
    "vendors": [],
    "frameworks": [
      "sglang",
      "trt",
      "vllm"
    ],
    "deployment": [
      "single-node"
    ],
    "spec": [],
    "optimal": false,
    "best": false,
    "format": "json",
    "tcoBasis": "internal",
    "power": [],
    "allPoints": true,
    "userCosts": {},
    "userPowers": {},
    "priceSource": "normalized",
    "dates": [],
    "unofficialrun": null
  },
  "metric": {
    "key": "tpPerGpu",
    "configKey": "y_tpPerGpu",
    "label": "Token Throughput per Chip (tok/s/chip)",
    "labelZh": "\u6bcf\u82af\u7247 token \u541e\u5410\u91cf\uff08tok/s/chip\uff09",
    "unit": "tok/s/chip",
    "polarity": "higher",
    "direction": "upper_left"
  },
  "xAxis": {
    "mode": "interactivity",
    "field": "median_intvty",
    "label": "Interactivity (tok/s/user)"
  },
  "frontier": {
    "direction": "upper_left",
    "points": 102
  },
  "hardware": [
    {
      "key": "b300_sglang",
      "label": "B300 (SGLang)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b300_sglang_mtp",
      "label": "B300 (SGLang, MTP)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b300_trt",
      "label": "B300 (TRTLLM)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b300_trt_mtp",
      "label": "B300 (TRTLLM, MTP)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b300_vllm",
      "label": "B300 (vLLM)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b300_vllm_mtp",
      "label": "B300 (vLLM, MTP)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b200_sglang",
      "label": "B200 (SGLang)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b200_trt",
      "label": "B200 (TRTLLM)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b200_trt_mtp",
      "label": "B200 (TRTLLM, MTP)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b200_vllm",
      "label": "B200 (vLLM)",
      "vendor": "NVIDIA"
    },
    {
      "key": "b200_vllm_mtp",
      "label": "B200 (vLLM, MTP)",
      "vendor": "NVIDIA"
    }
  ],
  "series": [
    {
      "hwKey": "b300_sglang",
      "gpu": "b300",
      "framework": "sglang",
      "specMethod": "none",
      "label": "B300 (SGLang)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 429921,
          "precision": "fp4",
          "x": 3.2867501036848155,
          "y": 10672.443754896221,
          "concurrency": 4096,
          "tp": 64,
          "date": "2026-06-26",
          "runId": 28078823761,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 10672.443754896221,
            "outputTputPerGpu": 1185.5670221305759,
            "inputTputPerGpu": 9486.876732765646
          }
        },
        {
          "id": 429928,
          "precision": "fp4",
          "x": 4.012409487480614,
          "y": 9133.539513004614,
          "concurrency": 2048,
          "tp": 64,
          "date": "2026-06-26",
          "runId": 28078823761,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 9133.539513004614,
            "outputTputPerGpu": 1015.7595985902257,
            "inputTputPerGpu": 8117.779914414389
          }
        },
        {
          "id": 429927,
          "precision": "fp4",
          "x": 5.886861101110446,
          "y": 6623.059935808474,
          "concurrency": 512,
          "tp": 4,
          "date": "2026-06-26",
          "runId": 28078823761,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 6623.059935808474,
            "outputTputPerGpu": 735.889608233771,
            "inputTputPerGpu": 5887.170327574702
          }
        },
        {
          "id": 429924,
          "precision": "fp4",
          "x": 33.21951910716134,
          "y": 2233.0895905704956,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-06-26",
          "runId": 28078823761,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2233.0895905704956,
            "outputTputPerGpu": 249.7215474709542,
            "inputTputPerGpu": 1983.3680430995414
          }
        },
        {
          "id": 429922,
          "precision": "fp4",
          "x": 101.93670064554867,
          "y": 98.5295619254958,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-06-26",
          "runId": 28078823761,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 98.5295619254958,
            "outputTputPerGpu": 11.00701900418953,
            "inputTputPerGpu": 87.52254292130627
          }
        }
      ]
    },
    {
      "hwKey": "b300_sglang_mtp",
      "gpu": "b300",
      "framework": "sglang",
      "specMethod": "mtp",
      "label": "B300 (SGLang, MTP)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 420345,
          "precision": "fp4",
          "x": 44.577426232264244,
          "y": 2624.8849559814266,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2624.8849559814266,
            "outputTputPerGpu": 293.4997047485705,
            "inputTputPerGpu": 2331.3852512328563
          }
        },
        {
          "id": 420353,
          "precision": "fp4",
          "x": 75.1903862664698,
          "y": 2225.8362202145327,
          "concurrency": 16,
          "tp": 4,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2225.8362202145327,
            "outputTputPerGpu": 246.60703348557087,
            "inputTputPerGpu": 1979.2291867289618
          }
        },
        {
          "id": 420354,
          "precision": "fp4",
          "x": 109.4494204060718,
          "y": 1539.2933623911692,
          "concurrency": 8,
          "tp": 4,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1539.2933623911692,
            "outputTputPerGpu": 173.02152801688655,
            "inputTputPerGpu": 1366.2718343742827
          }
        },
        {
          "id": 420349,
          "precision": "fp4",
          "x": 132.6737131454768,
          "y": 907.8732971352415,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 907.8732971352415,
            "outputTputPerGpu": 102.04788051060947,
            "inputTputPerGpu": 805.825416624632
          }
        },
        {
          "id": 420347,
          "precision": "fp4",
          "x": 140.28849139494895,
          "y": 1088.5106138004035,
          "concurrency": 4,
          "tp": 4,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1088.5106138004035,
            "outputTputPerGpu": 121.0735524590092,
            "inputTputPerGpu": 967.4370613413943
          }
        },
        {
          "id": 420340,
          "precision": "fp4",
          "x": 166.84883468471924,
          "y": 610.7714001338474,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 610.7714001338474,
            "outputTputPerGpu": 67.9352706505878,
            "inputTputPerGpu": 542.8361294832596
          }
        },
        {
          "id": 420342,
          "precision": "fp4",
          "x": 226.86900034343503,
          "y": 366.60934126776357,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 366.60934126776357,
            "outputTputPerGpu": 41.13411830443178,
            "inputTputPerGpu": 325.4752229633318
          }
        },
        {
          "id": 420355,
          "precision": "fp4",
          "x": 270.08681860870337,
          "y": 223.02425481079297,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-06-11",
          "runId": 27266692846,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 223.02425481079297,
            "outputTputPerGpu": 24.911685547448634,
            "inputTputPerGpu": 198.11256926334434
          }
        }
      ]
    },
    {
      "hwKey": "b300_trt",
      "gpu": "b300",
      "framework": "trt",
      "specMethod": "none",
      "label": "B300 (TRTLLM)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 420784,
          "precision": "fp4",
          "x": 20.82379911701535,
          "y": 5468.010965164718,
          "concurrency": 1024,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5468.010965164718,
            "outputTputPerGpu": 607.6814001483413,
            "inputTputPerGpu": 4860.329565016376
          }
        },
        {
          "id": 420771,
          "precision": "fp4",
          "x": 20.890719910739882,
          "y": 5428.728195747012,
          "concurrency": 512,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5428.728195747012,
            "outputTputPerGpu": 603.1871527504507,
            "inputTputPerGpu": 4825.541042996561
          }
        },
        {
          "id": 420767,
          "precision": "fp4",
          "x": 21.120789448301235,
          "y": 5182.285101344602,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5182.285101344602,
            "outputTputPerGpu": 575.9750573528831,
            "inputTputPerGpu": 4606.310043991718
          }
        },
        {
          "id": 420805,
          "precision": "fp4",
          "x": 26.146989274758578,
          "y": 3360.1271751229588,
          "concurrency": 64,
          "tp": 16,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3360.1271751229588,
            "outputTputPerGpu": 372.7767850616885,
            "inputTputPerGpu": 2987.3503900612704
          }
        },
        {
          "id": 420775,
          "precision": "fp4",
          "x": 31.977604259726885,
          "y": 2202.5530703107393,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2202.5530703107393,
            "outputTputPerGpu": 246.30671488839923,
            "inputTputPerGpu": 1956.24635542234
          }
        },
        {
          "id": 420799,
          "precision": "fp4",
          "x": 45.12870976978709,
          "y": 1553.7505273596873,
          "concurrency": 16,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1553.7505273596873,
            "outputTputPerGpu": 172.16549804773936,
            "inputTputPerGpu": 1381.585029311948
          }
        },
        {
          "id": 420804,
          "precision": "fp4",
          "x": 60.883760733436254,
          "y": 1038.838844668857,
          "concurrency": 8,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1038.838844668857,
            "outputTputPerGpu": 116.78300745476031,
            "inputTputPerGpu": 922.0558372140968
          }
        },
        {
          "id": 420788,
          "precision": "fp4",
          "x": 75.94552580443619,
          "y": 650.2028839415789,
          "concurrency": 4,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 650.2028839415789,
            "outputTputPerGpu": 72.32992441846069,
            "inputTputPerGpu": 577.8729595231182
          }
        },
        {
          "id": 420782,
          "precision": "fp4",
          "x": 88.53525444967875,
          "y": 381.36757878939926,
          "concurrency": 2,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 381.36757878939926,
            "outputTputPerGpu": 42.795148817592725,
            "inputTputPerGpu": 338.5724299718065
          }
        },
        {
          "id": 420778,
          "precision": "fp4",
          "x": 95.78907355460612,
          "y": 207.78266205689283,
          "concurrency": 1,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 207.78266205689283,
            "outputTputPerGpu": 23.211995114021736,
            "inputTputPerGpu": 184.5706669428711
          }
        }
      ]
    },
    {
      "hwKey": "b300_trt_mtp",
      "gpu": "b300",
      "framework": "trt",
      "specMethod": "mtp",
      "label": "B300 (TRTLLM, MTP)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": true,
      "points": [
        {
          "id": 420792,
          "precision": "fp4",
          "x": 44.763962948449546,
          "y": 2865.2230463788164,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": false,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 2865.2230463788164,
            "outputTputPerGpu": 320.37294291115313,
            "inputTputPerGpu": 2544.8501034676633
          }
        },
        {
          "id": 420770,
          "precision": "fp4",
          "x": 46.655374618718106,
          "y": 4256.553754340255,
          "concurrency": 64,
          "tp": 16,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": false,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4256.553754340255,
            "outputTputPerGpu": 472.17066109313697,
            "inputTputPerGpu": 3784.3830932471183
          }
        },
        {
          "id": 420802,
          "precision": "fp4",
          "x": 47.61512830783772,
          "y": 4986.60028664934,
          "concurrency": 1024,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4986.60028664934,
            "outputTputPerGpu": 554.1135594776839,
            "inputTputPerGpu": 4432.486727171656
          }
        },
        {
          "id": 420772,
          "precision": "fp4",
          "x": 47.93094557822679,
          "y": 4977.422813799249,
          "concurrency": 512,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4977.422813799249,
            "outputTputPerGpu": 553.0073081309871,
            "inputTputPerGpu": 4424.415505668261
          }
        },
        {
          "id": 420790,
          "precision": "fp4",
          "x": 48.364171758435496,
          "y": 4922.059847494746,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4922.059847494746,
            "outputTputPerGpu": 547.0065437229907,
            "inputTputPerGpu": 4375.053303771755
          }
        },
        {
          "id": 420795,
          "precision": "fp4",
          "x": 66.04824170096846,
          "y": 2143.693790200064,
          "concurrency": 16,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 2143.693790200064,
            "outputTputPerGpu": 237.50622867109453,
            "inputTputPerGpu": 1906.1875615289696
          }
        },
        {
          "id": 420807,
          "precision": "fp4",
          "x": 98.84514275408104,
          "y": 1503.020237952281,
          "concurrency": 8,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 1503.020237952281,
            "outputTputPerGpu": 168.94431208800484,
            "inputTputPerGpu": 1334.075925864276
          }
        },
        {
          "id": 420777,
          "precision": "fp4",
          "x": 134.277949046375,
          "y": 1069.2064916151194,
          "concurrency": 4,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 1069.2064916151194,
            "outputTputPerGpu": 118.92638125053105,
            "inputTputPerGpu": 950.2801103645884
          }
        },
        {
          "id": 420766,
          "precision": "fp4",
          "x": 172.88452720784375,
          "y": 687.3506719028771,
          "concurrency": 2,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 687.3506719028771,
            "outputTputPerGpu": 77.12177697630791,
            "inputTputPerGpu": 610.2288949265692
          }
        },
        {
          "id": 420789,
          "precision": "fp4",
          "x": 204.36651625216385,
          "y": 414.3572372315183,
          "concurrency": 1,
          "tp": 4,
          "date": "2026-06-12",
          "runId": 27370207913,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 414.3572372315183,
            "outputTputPerGpu": 46.28347354855335,
            "inputTputPerGpu": 368.07376368296497
          }
        }
      ]
    },
    {
      "hwKey": "b300_vllm",
      "gpu": "b300",
      "framework": "vllm",
      "specMethod": "none",
      "label": "B300 (vLLM)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 434576,
          "precision": "fp4",
          "x": 5.19782130080069,
          "y": 10487.690602618402,
          "concurrency": 2048,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 10487.690602618402,
            "outputTputPerGpu": 1166.3575092094495,
            "inputTputPerGpu": 9321.333093408954
          }
        },
        {
          "id": 434547,
          "precision": "fp4",
          "x": 6.354245835808449,
          "y": 7153.048219869655,
          "concurrency": 512,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 7153.048219869655,
            "outputTputPerGpu": 794.7767200078339,
            "inputTputPerGpu": 6358.271499861821
          }
        },
        {
          "id": 434546,
          "precision": "fp4",
          "x": 8.518467155453871,
          "y": 4970.785632612325,
          "concurrency": 256,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 4970.785632612325,
            "outputTputPerGpu": 552.4683578466035,
            "inputTputPerGpu": 4418.317274765721
          }
        },
        {
          "id": 434573,
          "precision": "fp4",
          "x": 23.679747067421005,
          "y": 3331.776744038019,
          "concurrency": 64,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3331.776744038019,
            "outputTputPerGpu": 369.6315521570529,
            "inputTputPerGpu": 2962.145191880966
          }
        },
        {
          "id": 434557,
          "precision": "fp4",
          "x": 36.52468014297694,
          "y": 2505.048699993063,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2505.048699993063,
            "outputTputPerGpu": 280.13414262190645,
            "inputTputPerGpu": 2224.9145573711567
          }
        },
        {
          "id": 434608,
          "precision": "fp4",
          "x": 50.93477393102296,
          "y": 1745.679659168149,
          "concurrency": 16,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1745.679659168149,
            "outputTputPerGpu": 193.43247365663942,
            "inputTputPerGpu": 1552.2471855115098
          }
        },
        {
          "id": 434592,
          "precision": "fp4",
          "x": 64.90171088547042,
          "y": 1053.8347266362105,
          "concurrency": 8,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1053.8347266362105,
            "outputTputPerGpu": 118.46879751216078,
            "inputTputPerGpu": 935.3659291240497
          }
        },
        {
          "id": 434545,
          "precision": "fp4",
          "x": 76.37433501822727,
          "y": 653.6457322120638,
          "concurrency": 4,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 653.6457322120638,
            "outputTputPerGpu": 72.71291403806806,
            "inputTputPerGpu": 580.9328181739957
          }
        },
        {
          "id": 434509,
          "precision": "fp4",
          "x": 81.09552392544063,
          "y": 347.4175751715158,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 347.4175751715158,
            "outputTputPerGpu": 38.64745539341294,
            "inputTputPerGpu": 308.77011977810287
          }
        },
        {
          "id": 434587,
          "precision": "fp4",
          "x": 85.81795251177519,
          "y": 370.6924772319031,
          "concurrency": 2,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 370.6924772319031,
            "outputTputPerGpu": 41.59724268921613,
            "inputTputPerGpu": 329.09523454268697
          }
        },
        {
          "id": 434589,
          "precision": "fp4",
          "x": 92.93533585767669,
          "y": 197.91242493465393,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 197.91242493465393,
            "outputTputPerGpu": 22.208735479861907,
            "inputTputPerGpu": 175.70368945479203
          }
        },
        {
          "id": 434564,
          "precision": "fp4",
          "x": 93.41855048414203,
          "y": 202.81356864613872,
          "concurrency": 1,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 202.81356864613872,
            "outputTputPerGpu": 22.65688348521817,
            "inputTputPerGpu": 180.15668516092055
          }
        },
        {
          "id": 434559,
          "precision": "fp4",
          "x": 98.58970293280132,
          "y": 105.60211329656138,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 105.60211329656138,
            "outputTputPerGpu": 11.797113934361768,
            "inputTputPerGpu": 93.80499936219962
          }
        }
      ]
    },
    {
      "hwKey": "b300_vllm_mtp",
      "gpu": "b300",
      "framework": "vllm",
      "specMethod": "mtp",
      "label": "B300 (vLLM, MTP)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 434522,
          "precision": "fp4",
          "x": 9.535946916556142,
          "y": 5211.145891806475,
          "concurrency": 256,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5211.145891806475,
            "outputTputPerGpu": 579.1337349472091,
            "inputTputPerGpu": 4632.012156859267
          }
        },
        {
          "id": 434558,
          "precision": "fp4",
          "x": 10.72725937173371,
          "y": 9254.298877567646,
          "concurrency": 512,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 9254.298877567646,
            "outputTputPerGpu": 1028.1816719960307,
            "inputTputPerGpu": 8226.117205571616
          }
        },
        {
          "id": 434513,
          "precision": "fp4",
          "x": 11.412325660036636,
          "y": 6237.0553856144825,
          "concurrency": 256,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 6237.0553856144825,
            "outputTputPerGpu": 693.1468156020801,
            "inputTputPerGpu": 5543.908570012402
          }
        },
        {
          "id": 434503,
          "precision": "fp4",
          "x": 16.383379752253514,
          "y": 4504.9676442578375,
          "concurrency": 128,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 4504.9676442578375,
            "outputTputPerGpu": 498.934930316066,
            "inputTputPerGpu": 4006.0327139417714
          }
        },
        {
          "id": 434551,
          "precision": "fp4",
          "x": 26.27425616508162,
          "y": 3588.0954331078474,
          "concurrency": 64,
          "tp": 16,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3588.0954331078474,
            "outputTputPerGpu": 398.0199688511603,
            "inputTputPerGpu": 3190.075464256687
          }
        },
        {
          "id": 434514,
          "precision": "fp4",
          "x": 26.83248805475358,
          "y": 3654.4604934517615,
          "concurrency": 64,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3654.4604934517615,
            "outputTputPerGpu": 405.3817070611195,
            "inputTputPerGpu": 3249.078786390642
          }
        },
        {
          "id": 434598,
          "precision": "fp4",
          "x": 42.3867614807639,
          "y": 2820.4948699194565,
          "concurrency": 32,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2820.4948699194565,
            "outputTputPerGpu": 315.37169264498453,
            "inputTputPerGpu": 2505.123177274472
          }
        },
        {
          "id": 434561,
          "precision": "fp4",
          "x": 63.07544775437514,
          "y": 2132.8319949051397,
          "concurrency": 16,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2132.8319949051397,
            "outputTputPerGpu": 236.302817974619,
            "inputTputPerGpu": 1896.5291769305206
          }
        },
        {
          "id": 434548,
          "precision": "fp4",
          "x": 93.03108290335881,
          "y": 1459.9343822392123,
          "concurrency": 8,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1459.9343822392123,
            "outputTputPerGpu": 164.101323237712,
            "inputTputPerGpu": 1295.8330590015003
          }
        },
        {
          "id": 434569,
          "precision": "fp4",
          "x": 96.90573199161597,
          "y": 756.3393520170985,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 756.3393520170985,
            "outputTputPerGpu": 85.01497738027987,
            "inputTputPerGpu": 671.3243746368186
          }
        },
        {
          "id": 434606,
          "precision": "fp4",
          "x": 106.8387064340745,
          "y": 905.9849903069776,
          "concurrency": 4,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 905.9849903069776,
            "outputTputPerGpu": 100.7714760520658,
            "inputTputPerGpu": 805.2135142549118
          }
        },
        {
          "id": 434607,
          "precision": "fp4",
          "x": 120.9655404002265,
          "y": 508.8340434517695,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 508.8340434517695,
            "outputTputPerGpu": 56.59691735820238,
            "inputTputPerGpu": 452.2371260935671
          }
        },
        {
          "id": 434562,
          "precision": "fp4",
          "x": 154.50999528433834,
          "y": 630.4194130451032,
          "concurrency": 2,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 630.4194130451032,
            "outputTputPerGpu": 70.73400428896254,
            "inputTputPerGpu": 559.6854087561406
          }
        },
        {
          "id": 434542,
          "precision": "fp4",
          "x": 172.92142251449587,
          "y": 340.1430061221914,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 340.1430061221914,
            "outputTputPerGpu": 38.16455577992544,
            "inputTputPerGpu": 301.978450342266
          }
        },
        {
          "id": 434597,
          "precision": "fp4",
          "x": 198.7559207304595,
          "y": 200.70883903107944,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 200.70883903107944,
            "outputTputPerGpu": 22.41906598355224,
            "inputTputPerGpu": 178.2897730475272
          }
        },
        {
          "id": 434501,
          "precision": "fp4",
          "x": 198.92806055405083,
          "y": 404.1402119084636,
          "concurrency": 1,
          "tp": 4,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 404.1402119084636,
            "outputTputPerGpu": 45.142237487506144,
            "inputTputPerGpu": 358.99797442095746
          }
        }
      ]
    },
    {
      "hwKey": "b200_sglang",
      "gpu": "b200",
      "framework": "sglang",
      "specMethod": "none",
      "label": "B200 (SGLang)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 431230,
          "precision": "fp4",
          "x": 8.016014510445743,
          "y": 8168.662307275478,
          "concurrency": 1024,
          "tp": 64,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 8168.662307275478,
            "outputTputPerGpu": 907.8153244110418,
            "inputTputPerGpu": 7260.8469828644365
          }
        },
        {
          "id": 431220,
          "precision": "fp4",
          "x": 12.306194969728438,
          "y": 6821.348582333877,
          "concurrency": 512,
          "tp": 64,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 6821.348582333877,
            "outputTputPerGpu": 757.9215022258298,
            "inputTputPerGpu": 6063.427080108047
          }
        },
        {
          "id": 431216,
          "precision": "fp4",
          "x": 18.58296840182908,
          "y": 5125.220663810464,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5125.220663810464,
            "outputTputPerGpu": 569.6327407804108,
            "inputTputPerGpu": 4555.587923030053
          }
        },
        {
          "id": 431214,
          "precision": "fp4",
          "x": 25.567264371100762,
          "y": 3466.4945633769785,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3466.4945633769785,
            "outputTputPerGpu": 383.96800370530156,
            "inputTputPerGpu": 3082.5265596716768
          }
        },
        {
          "id": 431224,
          "precision": "fp4",
          "x": 31.842884222885765,
          "y": 2147.8212138920317,
          "concurrency": 64,
          "tp": 64,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2147.8212138920317,
            "outputTputPerGpu": 238.28201888598633,
            "inputTputPerGpu": 1909.5391950060452
          }
        },
        {
          "id": 431217,
          "precision": "fp4",
          "x": 39.46403928682314,
          "y": 1320.7860914577002,
          "concurrency": 32,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1320.7860914577002,
            "outputTputPerGpu": 147.70063325254569,
            "inputTputPerGpu": 1173.0854582051545
          }
        },
        {
          "id": 431228,
          "precision": "fp4",
          "x": 53.82929412050623,
          "y": 897.2678503263065,
          "concurrency": 16,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 897.2678503263065,
            "outputTputPerGpu": 99.42301779691807,
            "inputTputPerGpu": 797.8448325293884
          }
        },
        {
          "id": 431211,
          "precision": "fp4",
          "x": 67.938637079146,
          "y": 557.27426705698,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 557.27426705698,
            "outputTputPerGpu": 62.64702674340842,
            "inputTputPerGpu": 494.6272403135715
          }
        },
        {
          "id": 431219,
          "precision": "fp4",
          "x": 79.01523764055041,
          "y": 327.05914480184566,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 327.05914480184566,
            "outputTputPerGpu": 36.382741153774106,
            "inputTputPerGpu": 290.6764036480715
          }
        },
        {
          "id": 431215,
          "precision": "fp4",
          "x": 90.05761238164762,
          "y": 183.84722865080107,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 183.84722865080107,
            "outputTputPerGpu": 20.630409996539868,
            "inputTputPerGpu": 163.2168186542612
          }
        },
        {
          "id": 431213,
          "precision": "fp4",
          "x": 101.59682696975959,
          "y": 103.30081791850719,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-07-01",
          "runId": 28403258484,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 103.30081791850719,
            "outputTputPerGpu": 11.540029649549359,
            "inputTputPerGpu": 91.76078826895784
          }
        }
      ]
    },
    {
      "hwKey": "b200_trt",
      "gpu": "b200",
      "framework": "trt",
      "specMethod": "none",
      "label": "B200 (TRTLLM)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 422367,
          "precision": "fp4",
          "x": 20.153376439869287,
          "y": 5231.265775926892,
          "concurrency": 512,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5231.265775926892,
            "outputTputPerGpu": 581.247060984607,
            "inputTputPerGpu": 4650.018714942285
          }
        },
        {
          "id": 422365,
          "precision": "fp4",
          "x": 20.530801053298962,
          "y": 5025.503080615016,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 5025.503080615016,
            "outputTputPerGpu": 558.549822805638,
            "inputTputPerGpu": 4466.953257809379
          }
        },
        {
          "id": 422364,
          "precision": "fp4",
          "x": 26.877336497552818,
          "y": 3407.428414557078,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3407.428414557078,
            "outputTputPerGpu": 377.4255121957106,
            "inputTputPerGpu": 3030.0029023613674
          }
        },
        {
          "id": 422368,
          "precision": "fp4",
          "x": 33.25043067037036,
          "y": 2195.3589265186793,
          "concurrency": 64,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2195.3589265186793,
            "outputTputPerGpu": 243.55591322348252,
            "inputTputPerGpu": 1951.8030132951967
          }
        },
        {
          "id": 422357,
          "precision": "fp4",
          "x": 38.26473292915482,
          "y": 1315.1261007079572,
          "concurrency": 32,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1315.1261007079572,
            "outputTputPerGpu": 147.06768880881827,
            "inputTputPerGpu": 1168.058411899139
          }
        },
        {
          "id": 422366,
          "precision": "fp4",
          "x": 40.39096440226382,
          "y": 1342.5672268987971,
          "concurrency": 32,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1342.5672268987971,
            "outputTputPerGpu": 150.13637021132823,
            "inputTputPerGpu": 1192.4308566874688
          }
        },
        {
          "id": 422354,
          "precision": "fp4",
          "x": 52.769049071643515,
          "y": 904.7466050528406,
          "concurrency": 16,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 904.7466050528406,
            "outputTputPerGpu": 100.25171166353167,
            "inputTputPerGpu": 804.4948933893089
          }
        },
        {
          "id": 422359,
          "precision": "fp4",
          "x": 69.36153863980167,
          "y": 589.7350233414081,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 589.7350233414081,
            "outputTputPerGpu": 66.29616324095623,
            "inputTputPerGpu": 523.4388601004518
          }
        },
        {
          "id": 422358,
          "precision": "fp4",
          "x": 82.73167616904085,
          "y": 353.22469481544863,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 353.22469481544863,
            "outputTputPerGpu": 39.29345148987498,
            "inputTputPerGpu": 313.93124332557363
          }
        },
        {
          "id": 422356,
          "precision": "fp4",
          "x": 92.63238503051218,
          "y": 199.47914670726647,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 199.47914670726647,
            "outputTputPerGpu": 22.384545105912323,
            "inputTputPerGpu": 177.09460160135416
          }
        },
        {
          "id": 422355,
          "precision": "fp4",
          "x": 96.05924719960281,
          "y": 104.23457647914503,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 104.23457647914503,
            "outputTputPerGpu": 11.644342487457205,
            "inputTputPerGpu": 92.59023399168782
          }
        }
      ]
    },
    {
      "hwKey": "b200_trt_mtp",
      "gpu": "b200",
      "framework": "trt",
      "specMethod": "mtp",
      "label": "B200 (TRTLLM, MTP)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": true,
      "points": [
        {
          "id": 422361,
          "precision": "fp4",
          "x": 46.3126309751715,
          "y": 4743.441062548816,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4743.441062548816,
            "outputTputPerGpu": 527.1559837492017,
            "inputTputPerGpu": 4216.285078799615
          }
        },
        {
          "id": 422360,
          "precision": "fp4",
          "x": 46.7019539657051,
          "y": 4647.723310861159,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 4647.723310861159,
            "outputTputPerGpu": 514.7454297898498,
            "inputTputPerGpu": 4132.977881071309
          }
        },
        {
          "id": 422351,
          "precision": "fp4",
          "x": 52.15764847639465,
          "y": 1701.1543327260865,
          "concurrency": 32,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": false,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 1701.1543327260865,
            "outputTputPerGpu": 190.21340087652615,
            "inputTputPerGpu": 1510.9409318495605
          }
        },
        {
          "id": 422363,
          "precision": "fp4",
          "x": 65.48187994492142,
          "y": 3436.9316885853464,
          "concurrency": 64,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 3436.9316885853464,
            "outputTputPerGpu": 381.25168885194154,
            "inputTputPerGpu": 3055.6799997334047
          }
        },
        {
          "id": 422362,
          "precision": "fp4",
          "x": 75.43866542940596,
          "y": 2225.5275406677847,
          "concurrency": 32,
          "tp": 64,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 2225.5275406677847,
            "outputTputPerGpu": 248.84583021718873,
            "inputTputPerGpu": 1976.681710450596
          }
        },
        {
          "id": 422348,
          "precision": "fp4",
          "x": 80.56002475558847,
          "y": 1294.3852261635939,
          "concurrency": 16,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 1294.3852261635939,
            "outputTputPerGpu": 143.40879976379742,
            "inputTputPerGpu": 1150.9764263997965
          }
        },
        {
          "id": 422353,
          "precision": "fp4",
          "x": 114.00934737595382,
          "y": 900.8407144670141,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 900.8407144670141,
            "outputTputPerGpu": 101.2573955849345,
            "inputTputPerGpu": 799.5833188820795
          }
        },
        {
          "id": 422352,
          "precision": "fp4",
          "x": 153.23185078292815,
          "y": 623.1832366287983,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 623.1832366287983,
            "outputTputPerGpu": 69.31582231258533,
            "inputTputPerGpu": 553.8674143162131
          }
        },
        {
          "id": 422350,
          "precision": "fp4",
          "x": 186.65349779609048,
          "y": 376.80994316665897,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 376.80994316665897,
            "outputTputPerGpu": 42.278641146742864,
            "inputTputPerGpu": 334.5313020199161
          }
        },
        {
          "id": 422349,
          "precision": "fp4",
          "x": 226.3558998441542,
          "y": 236.0453078458435,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-06-12",
          "runId": 27434759052,
          "frontier": true,
          "bestPerSku": true,
          "metrics": {
            "tpPerGpu": 236.0453078458435,
            "outputTputPerGpu": 26.366129948489327,
            "inputTputPerGpu": 209.67917789735418
          }
        }
      ]
    },
    {
      "hwKey": "b200_vllm",
      "gpu": "b200",
      "framework": "vllm",
      "specMethod": "none",
      "label": "B200 (vLLM)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 434532,
          "precision": "fp4",
          "x": 6.040621980084848,
          "y": 6850.500611898631,
          "concurrency": 1024,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 6850.500611898631,
            "outputTputPerGpu": 761.3228704325074,
            "inputTputPerGpu": 6089.177741466124
          }
        },
        {
          "id": 434553,
          "precision": "fp4",
          "x": 8.161189088187614,
          "y": 4631.595102382103,
          "concurrency": 512,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 4631.595102382103,
            "outputTputPerGpu": 514.6175239879303,
            "inputTputPerGpu": 4116.977578394172
          }
        },
        {
          "id": 434599,
          "precision": "fp4",
          "x": 10.469895940754094,
          "y": 3031.9305152721668,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 3031.9305152721668,
            "outputTputPerGpu": 336.9780547138832,
            "inputTputPerGpu": 2694.9524605582837
          }
        },
        {
          "id": 434544,
          "precision": "fp4",
          "x": 10.926833967216396,
          "y": 2749.021708629016,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2749.021708629016,
            "outputTputPerGpu": 305.5347024194203,
            "inputTputPerGpu": 2443.4870062095956
          }
        },
        {
          "id": 434543,
          "precision": "fp4",
          "x": 13.48055390973286,
          "y": 1960.5268084466582,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1960.5268084466582,
            "outputTputPerGpu": 217.1587322833269,
            "inputTputPerGpu": 1743.3680761633314
          }
        },
        {
          "id": 434516,
          "precision": "fp4",
          "x": 17.34628939964363,
          "y": 2368.192585251419,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2368.192585251419,
            "outputTputPerGpu": 262.31403590111387,
            "inputTputPerGpu": 2105.878549350305
          }
        },
        {
          "id": 434600,
          "precision": "fp4",
          "x": 18.53047418976194,
          "y": 1307.430186867052,
          "concurrency": 64,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1307.430186867052,
            "outputTputPerGpu": 145.04796882727135,
            "inputTputPerGpu": 1162.3822180397806
          }
        },
        {
          "id": 434574,
          "precision": "fp4",
          "x": 25.63488618305244,
          "y": 1757.0537278831694,
          "concurrency": 64,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1757.0537278831694,
            "outputTputPerGpu": 194.92977667935276,
            "inputTputPerGpu": 1562.1239512038167
          }
        },
        {
          "id": 434538,
          "precision": "fp4",
          "x": 37.905137047345555,
          "y": 1290.4413216333076,
          "concurrency": 32,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1290.4413216333076,
            "outputTputPerGpu": 144.30724370373616,
            "inputTputPerGpu": 1146.1340779295715
          }
        },
        {
          "id": 434556,
          "precision": "fp4",
          "x": 55.87642167745222,
          "y": 943.6503038333353,
          "concurrency": 16,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 943.6503038333353,
            "outputTputPerGpu": 104.56249036223623,
            "inputTputPerGpu": 839.0878134710991
          }
        },
        {
          "id": 434505,
          "precision": "fp4",
          "x": 69.23666886716083,
          "y": 582.1186907650213,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 582.1186907650213,
            "outputTputPerGpu": 65.43995900041337,
            "inputTputPerGpu": 516.6787317646078
          }
        },
        {
          "id": 434578,
          "precision": "fp4",
          "x": 79.82587930818637,
          "y": 341.25765872553023,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 341.25765872553023,
            "outputTputPerGpu": 37.962213445144066,
            "inputTputPerGpu": 303.29544528038616
          }
        },
        {
          "id": 434536,
          "precision": "fp4",
          "x": 90.78269983782113,
          "y": 192.35360945985383,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 192.35360945985383,
            "outputTputPerGpu": 21.584953205949798,
            "inputTputPerGpu": 170.76865625390403
          }
        },
        {
          "id": 434507,
          "precision": "fp4",
          "x": 96.26089241167273,
          "y": 102.32814157277487,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 102.32814157277487,
            "outputTputPerGpu": 11.431369194624196,
            "inputTputPerGpu": 90.89677237815067
          }
        }
      ]
    },
    {
      "hwKey": "b200_vllm_mtp",
      "gpu": "b200",
      "framework": "vllm",
      "specMethod": "mtp",
      "label": "B200 (vLLM, MTP)",
      "vendor": "NVIDIA",
      "deployment": "single-node",
      "kvOffload": false,
      "bestPerSku": false,
      "points": [
        {
          "id": 434582,
          "precision": "fp4",
          "x": 9.502416482625236,
          "y": 2738.7342705812853,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2738.7342705812853,
            "outputTputPerGpu": 304.36557334606323,
            "inputTputPerGpu": 2434.368697235222
          }
        },
        {
          "id": 434555,
          "precision": "fp4",
          "x": 10.948717184401247,
          "y": 2990.4721069011007,
          "concurrency": 256,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2990.4721069011007,
            "outputTputPerGpu": 332.3421213841158,
            "inputTputPerGpu": 2658.129985516985
          }
        },
        {
          "id": 434517,
          "precision": "fp4",
          "x": 14.708743467939756,
          "y": 2113.605045916348,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2113.605045916348,
            "outputTputPerGpu": 234.0863396974938,
            "inputTputPerGpu": 1879.5187062188543
          }
        },
        {
          "id": 434529,
          "precision": "fp4",
          "x": 19.917932644517734,
          "y": 2730.5070227993983,
          "concurrency": 128,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2730.5070227993983,
            "outputTputPerGpu": 302.40957066238457,
            "inputTputPerGpu": 2428.0974521370135
          }
        },
        {
          "id": 434527,
          "precision": "fp4",
          "x": 23.116844051547897,
          "y": 1610.6960173479981,
          "concurrency": 64,
          "tp": 64,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": false,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1610.6960173479981,
            "outputTputPerGpu": 178.67116151318626,
            "inputTputPerGpu": 1432.024855834812
          }
        },
        {
          "id": 434508,
          "precision": "fp4",
          "x": 31.232207174402674,
          "y": 2116.9075864951496,
          "concurrency": 64,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 2116.9075864951496,
            "outputTputPerGpu": 234.8241587620725,
            "inputTputPerGpu": 1882.083427733077
          }
        },
        {
          "id": 434590,
          "precision": "fp4",
          "x": 48.57844447984372,
          "y": 1618.7548035302677,
          "concurrency": 32,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1618.7548035302677,
            "outputTputPerGpu": 180.99995423183245,
            "inputTputPerGpu": 1437.7548492984354
          }
        },
        {
          "id": 434520,
          "precision": "fp4",
          "x": 69.13618158130294,
          "y": 1135.8391029919178,
          "concurrency": 16,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 1135.8391029919178,
            "outputTputPerGpu": 125.84300190727922,
            "inputTputPerGpu": 1009.9961010846386
          }
        },
        {
          "id": 434526,
          "precision": "fp4",
          "x": 94.96000141330688,
          "y": 687.8280329128703,
          "concurrency": 8,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 687.8280329128703,
            "outputTputPerGpu": 77.31408461513995,
            "inputTputPerGpu": 610.5139482977304
          }
        },
        {
          "id": 434539,
          "precision": "fp4",
          "x": 121.90739992804072,
          "y": 495.28916859251797,
          "concurrency": 4,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 495.28916859251797,
            "outputTputPerGpu": 55.09033938272752,
            "inputTputPerGpu": 440.19882920979046
          }
        },
        {
          "id": 434535,
          "precision": "fp4",
          "x": 164.91184561302043,
          "y": 328.24862445156396,
          "concurrency": 2,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 328.24862445156396,
            "outputTputPerGpu": 36.82998830516951,
            "inputTputPerGpu": 291.4186361463944
          }
        },
        {
          "id": 434583,
          "precision": "fp4",
          "x": 206.57857559295388,
          "y": 205.5903204909616,
          "concurrency": 1,
          "tp": 8,
          "date": "2026-07-12",
          "runId": 29177379558,
          "frontier": true,
          "bestPerSku": false,
          "metrics": {
            "tpPerGpu": 205.5903204909616,
            "outputTputPerGpu": 22.96432475478971,
            "inputTputPerGpu": 182.6259957361719
          }
        }
      ]
    }
  ],
  "count": 120,
  "comparisons": [],
  "overlays": [],
  "pricing": {
    "source": "normalized",
    "inputPerMillion": 1,
    "cachedInputPerMillion": 0.1,
    "outputPerMillion": 1
  }
}
