{
  "source": "results/2026-08-29-nemotron35-llmd-pd",
  "combos": {
    "ISL1024/OSL8192": {
      "llmd": [
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 93.25,
          "tput_per_user": 96.26,
          "ttft_p50": 2727.25,
          "ttft_p95": 2747.91,
          "itl_p50": 10.39,
          "itl_p95": 10.43,
          "req_latency_p50": 87799.23,
          "req_latency_p95": 88175.3,
          "req_throughput": 0.01,
          "isl": 1024.0,
          "osl": 8190.0,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 1
        },
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 142.28,
          "tput_per_user": 75.24,
          "ttft_p50": 5100.27,
          "ttft_p95": 5106.12,
          "itl_p50": 13.29,
          "itl_p95": 13.42,
          "req_latency_p50": 113942.68,
          "req_latency_p95": 114981.76,
          "req_throughput": 0.02,
          "isl": 1024.0,
          "osl": 8189.0,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 2
        },
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 208.6,
          "tput_per_user": 56.8,
          "ttft_p50": 9933.96,
          "ttft_p95": 9939.73,
          "itl_p50": 17.78,
          "itl_p95": 17.96,
          "req_latency_p50": 155562.85,
          "req_latency_p95": 156987.73,
          "req_throughput": 0.03,
          "isl": 1024.0,
          "osl": 8190.25,
          "requests": 4.0,
          "errors": 0,
          "concurrency": 4
        },
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 292.48,
          "tput_per_user": 40.87,
          "ttft_p50": 17764.66,
          "ttft_p95": 20275.08,
          "itl_p50": 24.71,
          "itl_p95": 25.03,
          "req_latency_p50": 221700.49,
          "req_latency_p95": 223939.92,
          "req_throughput": 0.04,
          "isl": 1024.0,
          "osl": 8189.75,
          "requests": 8.0,
          "errors": 0,
          "concurrency": 8
        },
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 409.28,
          "tput_per_user": 30.45,
          "ttft_p50": 34475.23,
          "ttft_p95": 51764.73,
          "itl_p50": 32.8,
          "itl_p95": 33.87,
          "req_latency_p50": 306496.9,
          "req_latency_p95": 320008.04,
          "req_throughput": 0.05,
          "isl": 1024.0,
          "osl": 8188.0,
          "requests": 16.0,
          "errors": 0,
          "concurrency": 16
        },
        {
          "req_isl": 1024,
          "req_osl": 8192,
          "tput_total": 595.51,
          "tput_per_user": 23.47,
          "ttft_p50": 51833.94,
          "ttft_p95": 92812.35,
          "itl_p50": 42.81,
          "itl_p95": 44.17,
          "req_latency_p50": 406705.9,
          "req_latency_p95": 437315.44,
          "req_throughput": 0.07,
          "isl": 1024.0,
          "osl": 8189.59,
          "requests": 32.0,
          "errors": 0,
          "concurrency": 32
        }
      ]
    },
    "ISL8192/OSL1024": {
      "llmd": [
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 64.35,
          "tput_per_user": 90.27,
          "ttft_p50": 4565.12,
          "ttft_p95": 4619.93,
          "itl_p50": 11.08,
          "itl_p95": 11.34,
          "req_latency_p50": 15893.1,
          "req_latency_p95": 16084.82,
          "req_throughput": 0.06,
          "isl": 8192.0,
          "osl": 1023.0,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 1
        },
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 92.24,
          "tput_per_user": 75.01,
          "ttft_p50": 6756.14,
          "ttft_p95": 8259.05,
          "itl_p50": 13.33,
          "itl_p95": 13.43,
          "req_latency_p50": 20382.47,
          "req_latency_p95": 21996.33,
          "req_throughput": 0.09,
          "isl": 8192.0,
          "osl": 1023.0,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 2
        },
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 128.8,
          "tput_per_user": 59.28,
          "ttft_p50": 9810.65,
          "ttft_p95": 14303.85,
          "itl_p50": 16.88,
          "itl_p95": 17.58,
          "req_latency_p50": 27613.09,
          "req_latency_p95": 31359.92,
          "req_throughput": 0.13,
          "isl": 8192.0,
          "osl": 1023.0,
          "requests": 4.0,
          "errors": 0,
          "concurrency": 4
        },
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 142.9,
          "tput_per_user": 60.53,
          "ttft_p50": 26955.62,
          "ttft_p95": 39056.06,
          "itl_p50": 16.79,
          "itl_p95": 18.95,
          "req_latency_p50": 44800.22,
          "req_latency_p95": 56501.39,
          "req_throughput": 0.14,
          "isl": 8192.0,
          "osl": 1024.0,
          "requests": 8.0,
          "errors": 0,
          "concurrency": 8
        },
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 197.07,
          "tput_per_user": 44.31,
          "ttft_p50": 33753.25,
          "ttft_p95": 60135.28,
          "itl_p50": 24.02,
          "itl_p95": 26.04,
          "req_latency_p50": 60395.34,
          "req_latency_p95": 81427.09,
          "req_throughput": 0.19,
          "isl": 8192.06,
          "osl": 1023.56,
          "requests": 16.0,
          "errors": 0,
          "concurrency": 16
        },
        {
          "req_isl": 8192,
          "req_osl": 1024,
          "tput_total": 220.97,
          "tput_per_user": 41.63,
          "ttft_p50": 66260.33,
          "ttft_p95": 123761.33,
          "itl_p50": 25.06,
          "itl_p95": 26.73,
          "req_latency_p50": 93429.2,
          "req_latency_p95": 145058.06,
          "req_throughput": 0.22,
          "isl": 8192.03,
          "osl": 1023.56,
          "requests": 32.0,
          "errors": 0,
          "concurrency": 32
        }
      ]
    },
    "ISL8192/OSL8192": {
      "llmd": [
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 90.89,
          "tput_per_user": 96.28,
          "ttft_p50": 5009.56,
          "ttft_p95": 5296.63,
          "itl_p50": 10.39,
          "itl_p95": 10.53,
          "req_latency_p50": 90084.28,
          "req_latency_p95": 90921.73,
          "req_throughput": 0.01,
          "isl": 8192.0,
          "osl": 8190.0,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 1
        },
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 136.35,
          "tput_per_user": 75.15,
          "ttft_p50": 7045.58,
          "ttft_p95": 8815.66,
          "itl_p50": 13.31,
          "itl_p95": 13.54,
          "req_latency_p50": 116054.7,
          "req_latency_p95": 119701.37,
          "req_throughput": 0.02,
          "isl": 8192.0,
          "osl": 8189.5,
          "requests": 2.0,
          "errors": 0,
          "concurrency": 2
        },
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 208.16,
          "tput_per_user": 57.49,
          "ttft_p50": 10346.53,
          "ttft_p95": 14861.72,
          "itl_p50": 17.4,
          "itl_p95": 17.46,
          "req_latency_p50": 152751.96,
          "req_latency_p95": 156902.44,
          "req_throughput": 0.03,
          "isl": 8192.0,
          "osl": 8182.5,
          "requests": 4.0,
          "errors": 0,
          "concurrency": 4
        },
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 277.21,
          "tput_per_user": 39.93,
          "ttft_p50": 18741.68,
          "ttft_p95": 31099.88,
          "itl_p50": 24.91,
          "itl_p95": 26.14,
          "req_latency_p50": 228163.46,
          "req_latency_p95": 235366.43,
          "req_throughput": 0.03,
          "isl": 8192.0,
          "osl": 8190.75,
          "requests": 8.0,
          "errors": 0,
          "concurrency": 8
        },
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 385.43,
          "tput_per_user": 29.75,
          "ttft_p50": 35982.91,
          "ttft_p95": 65115.68,
          "itl_p50": 34.11,
          "itl_p95": 36.09,
          "req_latency_p50": 320686.25,
          "req_latency_p95": 338602.68,
          "req_throughput": 0.05,
          "isl": 8192.06,
          "osl": 8190.25,
          "requests": 16.0,
          "errors": 0,
          "concurrency": 16
        },
        {
          "req_isl": 8192,
          "req_osl": 8192,
          "tput_total": 526.6,
          "tput_per_user": 22.9,
          "ttft_p50": 75411.94,
          "ttft_p95": 143421.66,
          "itl_p50": 43.86,
          "itl_p95": 46.83,
          "req_latency_p50": 450834.96,
          "req_latency_p95": 493459.13,
          "req_throughput": 0.06,
          "isl": 8192.03,
          "osl": 8190.62,
          "requests": 32.0,
          "errors": 0,
          "concurrency": 32
        }
      ]
    }
  }
}