{
  "schema": "aero.public.inference_benchmark.v1",
  "suite": "warm-cache",
  "publication_date": "2026-09-24",
  "scope": "Synthetic inference and mocked tools; not CFD/FEA",
  "rows": [
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 16.731716500013135,
      "wall_seconds_mean": 16.541558339993934,
      "wall_seconds_min": 14.931366299977526,
      "wall_seconds_max": 18.025823000003584,
      "wall_seconds_stdev": 1.133760985330175,
      "wall_seconds_p95": 18.025823000003584,
      "prompt_tokens_median": 15365.0,
      "prompt_tokens_mean": 15340.8,
      "prompt_tokens_min": 15244.0,
      "prompt_tokens_max": 15365.0,
      "prompt_tokens_stdev": 54.11284505549491,
      "prompt_tokens_p95": 15365.0,
      "output_tokens_median": 710.0,
      "output_tokens_mean": 691.8,
      "output_tokens_min": 619.0,
      "output_tokens_max": 710.0,
      "output_tokens_stdev": 40.69643719049617,
      "output_tokens_p95": 710.0,
      "prefill_tokens_per_second_median": 6504.554458128068,
      "prefill_tokens_per_second_mean": 6712.76221291863,
      "prefill_tokens_per_second_min": 5559.91994294252,
      "prefill_tokens_per_second_max": 8445.944074154104,
      "prefill_tokens_per_second_stdev": 1080.9062780646875,
      "prefill_tokens_per_second_p95": 8445.944074154104,
      "decode_tokens_per_second_median": 62.702923475690916,
      "decode_tokens_per_second_mean": 62.3548261167683,
      "decode_tokens_per_second_min": 60.901000277485544,
      "decode_tokens_per_second_max": 63.45766328148335,
      "decode_tokens_per_second_stdev": 1.113204245380119,
      "decode_tokens_per_second_p95": 63.45766328148335,
      "prefill_seconds_median": 2.362191,
      "prefill_seconds_mean": 2.3303184,
      "prefill_seconds_min": 1.8048899999999999,
      "prefill_seconds_max": 2.763529,
      "prefill_seconds_stdev": 0.35227479135300055,
      "prefill_seconds_p95": 2.763529,
      "decode_seconds_median": 11.229915,
      "decode_seconds_mean": 11.0991108,
      "decode_seconds_min": 9.871948,
      "decode_seconds_max": 11.658265,
      "decode_seconds_stdev": 0.7148734400421521,
      "decode_seconds_p95": 11.658265,
      "peak_gpu_memory_used_mib_median": 11438.0,
      "peak_system_memory_used_bytes_median": 44859408384.0,
      "gpu_energy_j_observed_median": 5604.693192667961
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 4096,
      "n": 5,
      "wall_seconds_median": 2.08706049999455,
      "wall_seconds_mean": 2.079625399992801,
      "wall_seconds_min": 1.9897436000173911,
      "wall_seconds_max": 2.1694802999845706,
      "wall_seconds_stdev": 0.06529933772452794,
      "wall_seconds_p95": 2.1694802999845706,
      "ttft_seconds_median": 0.8729013999691233,
      "ttft_seconds_mean": 0.8816336799878627,
      "ttft_seconds_min": 0.8198262000223622,
      "ttft_seconds_max": 0.9831509999930859,
      "ttft_seconds_stdev": 0.06117173179486046,
      "ttft_seconds_p95": 0.9831509999930859,
      "prompt_tokens_median": 3721.0,
      "prompt_tokens_mean": 3721.0,
      "prompt_tokens_min": 3721.0,
      "prompt_tokens_max": 3721.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 3721.0,
      "output_tokens_median": 79.0,
      "output_tokens_mean": 79.0,
      "output_tokens_min": 79.0,
      "output_tokens_max": 79.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 79.0,
      "prefill_tokens_per_second_median": 15000.826436285794,
      "prefill_tokens_per_second_mean": 15049.800326644061,
      "prefill_tokens_per_second_min": 14365.186909573831,
      "prefill_tokens_per_second_max": 15741.00317696678,
      "prefill_tokens_per_second_stdev": 632.7972762560886,
      "prefill_tokens_per_second_p95": 15741.00317696678,
      "decode_tokens_per_second_median": 65.78706309895374,
      "decode_tokens_per_second_mean": 65.98812307795565,
      "decode_tokens_per_second_min": 64.616866965322,
      "decode_tokens_per_second_max": 67.56144242824377,
      "decode_tokens_per_second_stdev": 1.1379744366608056,
      "decode_tokens_per_second_p95": 67.56144242824377,
      "load_seconds_median": 0.5967649,
      "load_seconds_mean": 0.59978946,
      "load_seconds_min": 0.5471606,
      "load_seconds_max": 0.6937586,
      "load_seconds_stdev": 0.05762971671492929,
      "load_seconds_p95": 0.6937586,
      "peak_gpu_memory_used_mib_median": 10892.0,
      "peak_system_memory_used_bytes_median": 44653137920.0,
      "gpu_energy_j_observed_median": 640.928237888813
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 1.9324110000161454,
      "wall_seconds_mean": 1.9366768199950457,
      "wall_seconds_min": 1.8876191999879666,
      "wall_seconds_max": 2.0035452999873087,
      "wall_seconds_stdev": 0.045701851464395586,
      "wall_seconds_p95": 2.0035452999873087,
      "ttft_seconds_median": 0.816715499968268,
      "ttft_seconds_mean": 0.8213770799920894,
      "ttft_seconds_min": 0.782776299980469,
      "ttft_seconds_max": 0.886808299983386,
      "ttft_seconds_stdev": 0.03956483425117247,
      "ttft_seconds_p95": 0.886808299983386,
      "prompt_tokens_median": 7791.0,
      "prompt_tokens_mean": 7791.0,
      "prompt_tokens_min": 7791.0,
      "prompt_tokens_max": 7791.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 7791.0,
      "output_tokens_median": 72.0,
      "output_tokens_mean": 72.0,
      "output_tokens_min": 72.0,
      "output_tokens_max": 72.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 72.0,
      "prefill_tokens_per_second_median": 35842.44230981561,
      "prefill_tokens_per_second_mean": 35823.696601157266,
      "prefill_tokens_per_second_min": 29757.653304814467,
      "prefill_tokens_per_second_max": 38949.74178485904,
      "prefill_tokens_per_second_stdev": 3732.501106823881,
      "prefill_tokens_per_second_p95": 38949.74178485904,
      "decode_tokens_per_second_median": 64.80519737682963,
      "decode_tokens_per_second_mean": 64.59600537248113,
      "decode_tokens_per_second_min": 63.22472437971716,
      "decode_tokens_per_second_max": 65.26373801685254,
      "decode_tokens_per_second_stdev": 0.8278051346727496,
      "decode_tokens_per_second_p95": 65.26373801685254,
      "load_seconds_median": 0.5433629,
      "load_seconds_mean": 0.54644388,
      "load_seconds_min": 0.530647,
      "load_seconds_max": 0.5691809,
      "load_seconds_stdev": 0.014162893545705975,
      "load_seconds_p95": 0.5691809,
      "peak_gpu_memory_used_mib_median": 11444.0,
      "peak_system_memory_used_bytes_median": 44177502208.0,
      "gpu_energy_j_observed_median": 587.7048515725136
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 16384,
      "n": 5,
      "wall_seconds_median": 2.0611767999944277,
      "wall_seconds_mean": 2.078317899990361,
      "wall_seconds_min": 2.0107550999964587,
      "wall_seconds_max": 2.168754899990745,
      "wall_seconds_stdev": 0.05936494447140963,
      "wall_seconds_p95": 2.168754899990745,
      "ttft_seconds_median": 0.9047881999867968,
      "ttft_seconds_mean": 0.9208425399963744,
      "ttft_seconds_min": 0.8440043000155129,
      "ttft_seconds_max": 0.9925371999852359,
      "ttft_seconds_stdev": 0.057653051016289196,
      "ttft_seconds_p95": 0.9925371999852359,
      "prompt_tokens_median": 15931.0,
      "prompt_tokens_mean": 15931.0,
      "prompt_tokens_min": 15931.0,
      "prompt_tokens_max": 15931.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15931.0,
      "output_tokens_median": 70.0,
      "output_tokens_mean": 70.0,
      "output_tokens_min": 70.0,
      "output_tokens_max": 70.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 70.0,
      "prefill_tokens_per_second_median": 59937.16962320585,
      "prefill_tokens_per_second_mean": 59773.39968910843,
      "prefill_tokens_per_second_min": 50204.04947608918,
      "prefill_tokens_per_second_max": 66922.91535391724,
      "prefill_tokens_per_second_stdev": 6241.932052387908,
      "prefill_tokens_per_second_p95": 66922.91535391724,
      "decode_tokens_per_second_median": 60.402783015082576,
      "decode_tokens_per_second_mean": 60.51956644102034,
      "decode_tokens_per_second_min": 59.544214556518945,
      "decode_tokens_per_second_max": 61.622159328494526,
      "decode_tokens_per_second_stdev": 0.8194799740384817,
      "decode_tokens_per_second_p95": 61.622159328494526,
      "load_seconds_median": 0.5251512,
      "load_seconds_mean": 0.52849554,
      "load_seconds_min": 0.4933514,
      "load_seconds_max": 0.5857867,
      "load_seconds_stdev": 0.037301017348498694,
      "load_seconds_p95": 0.5857867,
      "peak_gpu_memory_used_mib_median": 11953.0,
      "peak_system_memory_used_bytes_median": 44377329664.0,
      "gpu_energy_j_observed_median": 627.6881807780265
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 32768,
      "n": 5,
      "wall_seconds_median": 2.573112800018862,
      "wall_seconds_mean": 2.6040672800154425,
      "wall_seconds_min": 2.427987800037954,
      "wall_seconds_max": 2.835893999959808,
      "wall_seconds_stdev": 0.15835161302716536,
      "wall_seconds_p95": 2.835893999959808,
      "ttft_seconds_median": 1.2491398000274785,
      "ttft_seconds_mean": 1.2972310599987396,
      "ttft_seconds_min": 1.1641691999975592,
      "ttft_seconds_max": 1.4383806999539956,
      "ttft_seconds_stdev": 0.11387234215105652,
      "ttft_seconds_p95": 1.4383806999539956,
      "prompt_tokens_median": 32211.0,
      "prompt_tokens_mean": 32211.0,
      "prompt_tokens_min": 32211.0,
      "prompt_tokens_max": 32211.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32211.0,
      "output_tokens_median": 70.0,
      "output_tokens_mean": 70.0,
      "output_tokens_min": 70.0,
      "output_tokens_max": 70.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 70.0,
      "prefill_tokens_per_second_median": 99504.1935035448,
      "prefill_tokens_per_second_mean": 97388.37521606573,
      "prefill_tokens_per_second_min": 84291.09750353274,
      "prefill_tokens_per_second_max": 105239.59970333875,
      "prefill_tokens_per_second_stdev": 8405.78198888589,
      "prefill_tokens_per_second_p95": 105239.59970333875,
      "decode_tokens_per_second_median": 54.596690348631064,
      "decode_tokens_per_second_mean": 53.729550437947594,
      "decode_tokens_per_second_min": 50.1607652526347,
      "decode_tokens_per_second_max": 55.72786966684287,
      "decode_tokens_per_second_stdev": 2.319702526656472,
      "decode_tokens_per_second_p95": 55.72786966684287,
      "load_seconds_median": 0.6682103,
      "load_seconds_mean": 0.64524588,
      "load_seconds_min": 0.5253621,
      "load_seconds_max": 0.7144731,
      "load_seconds_stdev": 0.07914472530283996,
      "load_seconds_p95": 0.7144731,
      "peak_gpu_memory_used_mib_median": 12715.0,
      "peak_system_memory_used_bytes_median": 44686053376.0,
      "gpu_energy_j_observed_median": 740.1698657512665
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 18.084672899974976,
      "wall_seconds_mean": 18.023657119995914,
      "wall_seconds_min": 17.54265789996134,
      "wall_seconds_max": 18.346895300026517,
      "wall_seconds_stdev": 0.29388760045992646,
      "wall_seconds_p95": 18.346895300026517,
      "ttft_seconds_median": 1.1036828000214882,
      "ttft_seconds_mean": 1.2063489999971353,
      "ttft_seconds_min": 0.973166499985382,
      "ttft_seconds_max": 1.4751721000066027,
      "ttft_seconds_stdev": 0.21930417344819314,
      "ttft_seconds_p95": 1.4751721000066027,
      "prompt_tokens_median": 16309.0,
      "prompt_tokens_mean": 16309.0,
      "prompt_tokens_min": 16309.0,
      "prompt_tokens_max": 16309.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16309.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 49783.88014505672,
      "prefill_tokens_per_second_mean": 42732.361025761296,
      "prefill_tokens_per_second_min": 20051.73689917944,
      "prefill_tokens_per_second_max": 61391.42349504623,
      "prefill_tokens_per_second_stdev": 19503.096371205036,
      "prefill_tokens_per_second_p95": 61391.42349504623,
      "decode_tokens_per_second_median": 59.27286655318679,
      "decode_tokens_per_second_mean": 59.47242927419142,
      "decode_tokens_per_second_min": 58.80060409388622,
      "decode_tokens_per_second_max": 60.354029498173986,
      "decode_tokens_per_second_stdev": 0.7118944300388519,
      "decode_tokens_per_second_p95": 60.354029498173986,
      "load_seconds_median": 0.581086,
      "load_seconds_mean": 0.59934058,
      "load_seconds_min": 0.5394787,
      "load_seconds_max": 0.68993,
      "load_seconds_stdev": 0.05997523419255653,
      "load_seconds_p95": 0.68993,
      "peak_gpu_memory_used_mib_median": 12250.0,
      "peak_system_memory_used_bytes_median": 44899364864.0,
      "gpu_energy_j_observed_median": 6881.941198734045
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 2.6517274000216275,
      "wall_seconds_mean": 2.6119459000066856,
      "wall_seconds_min": 2.5110518999863416,
      "wall_seconds_max": 2.698217900004238,
      "wall_seconds_stdev": 0.0854880704214838,
      "wall_seconds_p95": 2.698217900004238,
      "ttft_seconds_median": 0.8987082000239752,
      "ttft_seconds_mean": 0.8826888200128451,
      "ttft_seconds_min": 0.8198618000023998,
      "ttft_seconds_max": 0.9428185999859124,
      "ttft_seconds_stdev": 0.05443670307561511,
      "ttft_seconds_p95": 0.9428185999859124,
      "prompt_tokens_median": 4113.0,
      "prompt_tokens_mean": 4113.0,
      "prompt_tokens_min": 4113.0,
      "prompt_tokens_max": 4113.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4113.0,
      "output_tokens_median": 111.0,
      "output_tokens_mean": 111.0,
      "output_tokens_min": 111.0,
      "output_tokens_max": 111.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 111.0,
      "prefill_tokens_per_second_median": 14953.264232504536,
      "prefill_tokens_per_second_mean": 15993.913788813796,
      "prefill_tokens_per_second_min": 14231.39071793612,
      "prefill_tokens_per_second_max": 19041.578511210595,
      "prefill_tokens_per_second_stdev": 2161.728653392486,
      "prefill_tokens_per_second_p95": 19041.578511210595,
      "decode_tokens_per_second_median": 64.32358821313709,
      "decode_tokens_per_second_mean": 64.22509330479484,
      "decode_tokens_per_second_min": 62.4377029225345,
      "decode_tokens_per_second_max": 65.64155453393906,
      "decode_tokens_per_second_stdev": 1.3580687339559216,
      "decode_tokens_per_second_p95": 65.64155453393906,
      "load_seconds_median": 0.5814766,
      "load_seconds_mean": 0.58499364,
      "load_seconds_min": 0.5526742,
      "load_seconds_max": 0.6201766,
      "load_seconds_stdev": 0.024534272619908655,
      "load_seconds_p95": 0.6201766,
      "peak_gpu_memory_used_mib_median": 20573.0,
      "peak_system_memory_used_bytes_median": 45347430400.0,
      "gpu_energy_j_observed_median": 826.763846719265
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 19.43443679995835,
      "wall_seconds_mean": 19.386967939976604,
      "wall_seconds_min": 19.077601699973457,
      "wall_seconds_max": 19.579395399952773,
      "wall_seconds_stdev": 0.19023796308109514,
      "wall_seconds_p95": 19.579395399952773,
      "ttft_seconds_median": 1.4358543999842368,
      "ttft_seconds_mean": 1.4563474999857136,
      "ttft_seconds_min": 1.3730071000172757,
      "ttft_seconds_max": 1.562482999986969,
      "ttft_seconds_stdev": 0.07243905927627176,
      "ttft_seconds_p95": 1.562482999986969,
      "prompt_tokens_median": 32566.0,
      "prompt_tokens_mean": 32566.0,
      "prompt_tokens_min": 32566.0,
      "prompt_tokens_max": 32566.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32566.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 50059.79629293728,
      "prefill_tokens_per_second_mean": 50934.67764884267,
      "prefill_tokens_per_second_min": 43432.2203432327,
      "prefill_tokens_per_second_max": 56764.46390473136,
      "prefill_tokens_per_second_stdev": 5320.854956264583,
      "prefill_tokens_per_second_p95": 56764.46390473136,
      "decode_tokens_per_second_median": 55.58249762732213,
      "decode_tokens_per_second_mean": 55.779726927518006,
      "decode_tokens_per_second_min": 55.12004290985101,
      "decode_tokens_per_second_max": 56.48708086797823,
      "decode_tokens_per_second_stdev": 0.5499873672590877,
      "decode_tokens_per_second_p95": 56.48708086797823,
      "load_seconds_median": 0.5613327,
      "load_seconds_mean": 0.55635998,
      "load_seconds_min": 0.5235459,
      "load_seconds_max": 0.5878585,
      "load_seconds_stdev": 0.0289061802195828,
      "load_seconds_p95": 0.5878585,
      "peak_gpu_memory_used_mib_median": 12968.0,
      "peak_system_memory_used_bytes_median": 44653490176.0,
      "gpu_energy_j_observed_median": 7413.12799840331
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 19.604493199964054,
      "wall_seconds_mean": 19.624288439983502,
      "wall_seconds_min": 19.470551599981263,
      "wall_seconds_max": 19.89395749999676,
      "wall_seconds_stdev": 0.16281148036687693,
      "wall_seconds_p95": 19.89395749999676,
      "ttft_seconds_median": 1.6604224999900907,
      "ttft_seconds_mean": 1.620401579979807,
      "ttft_seconds_min": 1.4692287999787368,
      "ttft_seconds_max": 1.752907999965828,
      "ttft_seconds_stdev": 0.11637875785130933,
      "ttft_seconds_p95": 1.752907999965828,
      "prompt_tokens_median": 32557.0,
      "prompt_tokens_mean": 32557.0,
      "prompt_tokens_min": 32557.0,
      "prompt_tokens_max": 32557.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32557.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 49735.86842611802,
      "prefill_tokens_per_second_mean": 48676.28792909166,
      "prefill_tokens_per_second_min": 42918.06455440063,
      "prefill_tokens_per_second_max": 53171.21369252252,
      "prefill_tokens_per_second_stdev": 3979.866513997183,
      "prefill_tokens_per_second_p95": 53171.21369252252,
      "decode_tokens_per_second_median": 55.75594241258639,
      "decode_tokens_per_second_mean": 55.55543163763573,
      "decode_tokens_per_second_min": 54.84922551248099,
      "decode_tokens_per_second_max": 56.06096697431617,
      "decode_tokens_per_second_stdev": 0.5313373032432939,
      "decode_tokens_per_second_p95": 56.06096697431617,
      "load_seconds_median": 0.674093,
      "load_seconds_mean": 0.65776336,
      "load_seconds_min": 0.5730311,
      "load_seconds_max": 0.7261291,
      "load_seconds_stdev": 0.05707941188088748,
      "load_seconds_p95": 0.7261291,
      "peak_gpu_memory_used_mib_median": 13015.0,
      "peak_system_memory_used_bytes_median": 45008302080.0,
      "gpu_energy_j_observed_median": 7458.223459795712
    },
    {
      "system_id": "3090",
      "model": "gemma3:4b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 9.864410399983171,
      "wall_seconds_mean": 10.02300574000692,
      "wall_seconds_min": 9.641889900027309,
      "wall_seconds_max": 10.57474509999156,
      "wall_seconds_stdev": 0.43148906826527467,
      "wall_seconds_p95": 10.57474509999156,
      "prompt_tokens_median": 15556.0,
      "prompt_tokens_mean": 15556.0,
      "prompt_tokens_min": 15556.0,
      "prompt_tokens_max": 15556.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15556.0,
      "output_tokens_median": 810.0,
      "output_tokens_mean": 810.0,
      "output_tokens_min": 810.0,
      "output_tokens_max": 810.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 810.0,
      "prefill_tokens_per_second_median": 18775.738790801868,
      "prefill_tokens_per_second_mean": 17729.37540532664,
      "prefill_tokens_per_second_min": 14342.151968173624,
      "prefill_tokens_per_second_max": 20240.686694832224,
      "prefill_tokens_per_second_stdev": 2525.041888644628,
      "prefill_tokens_per_second_p95": 20240.686694832224,
      "decode_tokens_per_second_median": 127.8439356992353,
      "decode_tokens_per_second_mean": 125.9479387937026,
      "decode_tokens_per_second_min": 121.61959462837781,
      "decode_tokens_per_second_max": 130.15778337117015,
      "decode_tokens_per_second_stdev": 3.9982642514910514,
      "decode_tokens_per_second_p95": 130.15778337117015,
      "prefill_seconds_median": 0.828516,
      "prefill_seconds_mean": 0.8927876,
      "prefill_seconds_min": 0.768551,
      "prefill_seconds_max": 1.084635,
      "prefill_seconds_stdev": 0.13532111072667116,
      "prefill_seconds_p95": 1.084635,
      "decode_seconds_median": 6.33585,
      "decode_seconds_mean": 6.436458,
      "decode_seconds_min": 6.223216,
      "decode_seconds_max": 6.660111000000001,
      "decode_seconds_stdev": 0.2059229373224364,
      "decode_seconds_p95": 6.660111000000001,
      "peak_gpu_memory_used_mib_median": 6007.0,
      "peak_system_memory_used_bytes_median": 44028944384.0,
      "gpu_energy_j_observed_median": 3047.999830610752
    },
    {
      "system_id": "3090",
      "model": "gemma3:4b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 3.3977833000244573,
      "wall_seconds_mean": 3.4074757199850865,
      "wall_seconds_min": 3.381815699976869,
      "wall_seconds_max": 3.452301599958446,
      "wall_seconds_stdev": 0.02904155930044744,
      "wall_seconds_p95": 3.452301599958446,
      "ttft_seconds_median": 0.7387341999565251,
      "ttft_seconds_mean": 0.7356045999913476,
      "ttft_seconds_min": 0.7044905999791808,
      "ttft_seconds_max": 0.7603728999965824,
      "ttft_seconds_stdev": 0.020084271998688008,
      "ttft_seconds_p95": 0.7603728999965824,
      "prompt_tokens_median": 16309.0,
      "prompt_tokens_mean": 16309.0,
      "prompt_tokens_min": 16309.0,
      "prompt_tokens_max": 16309.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16309.0,
      "output_tokens_median": 337.0,
      "output_tokens_mean": 337.0,
      "output_tokens_min": 337.0,
      "output_tokens_max": 337.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 337.0,
      "prefill_tokens_per_second_median": 221631.83214198353,
      "prefill_tokens_per_second_mean": 218418.4364909581,
      "prefill_tokens_per_second_min": 181442.9548867998,
      "prefill_tokens_per_second_max": 243381.58483808386,
      "prefill_tokens_per_second_stdev": 23019.726761810765,
      "prefill_tokens_per_second_p95": 243381.58483808386,
      "decode_tokens_per_second_median": 125.52388529575215,
      "decode_tokens_per_second_mean": 126.20502378773561,
      "decode_tokens_per_second_min": 124.2369290400688,
      "decode_tokens_per_second_max": 128.45259222667423,
      "decode_tokens_per_second_stdev": 1.762198539719259,
      "decode_tokens_per_second_p95": 128.45259222667423,
      "load_seconds_median": 0.5275578,
      "load_seconds_mean": 0.54066082,
      "load_seconds_min": 0.5209501,
      "load_seconds_max": 0.5849438,
      "load_seconds_stdev": 0.026027895344841092,
      "load_seconds_p95": 0.5849438,
      "peak_gpu_memory_used_mib_median": 6268.0,
      "peak_system_memory_used_bytes_median": 43074719744.0,
      "gpu_energy_j_observed_median": 1095.3904283249378
    },
    {
      "system_id": "3090",
      "model": "gemma3:4b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 3.623957200034056,
      "wall_seconds_mean": 3.6844438199885188,
      "wall_seconds_min": 3.6022194999968633,
      "wall_seconds_max": 3.8032898999517784,
      "wall_seconds_stdev": 0.09837096807481435,
      "wall_seconds_p95": 3.8032898999517784,
      "ttft_seconds_median": 0.6496930999564938,
      "ttft_seconds_mean": 0.6795930999913253,
      "ttft_seconds_min": 0.623220500012394,
      "ttft_seconds_max": 0.7652141999569722,
      "ttft_seconds_stdev": 0.06578602342049512,
      "ttft_seconds_p95": 0.7652141999569722,
      "prompt_tokens_median": 4113.0,
      "prompt_tokens_mean": 4113.0,
      "prompt_tokens_min": 4113.0,
      "prompt_tokens_max": 4113.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4113.0,
      "output_tokens_median": 387.0,
      "output_tokens_mean": 387.0,
      "output_tokens_min": 387.0,
      "output_tokens_max": 387.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 387.0,
      "prefill_tokens_per_second_median": 48400.76255030714,
      "prefill_tokens_per_second_mean": 46269.20282530687,
      "prefill_tokens_per_second_min": 38519.17061566989,
      "prefill_tokens_per_second_max": 54267.65710968321,
      "prefill_tokens_per_second_stdev": 6759.98597234267,
      "prefill_tokens_per_second_p95": 54267.65710968321,
      "decode_tokens_per_second_median": 128.9753657051503,
      "decode_tokens_per_second_mean": 128.8262773506386,
      "decode_tokens_per_second_min": 127.09639259714417,
      "decode_tokens_per_second_max": 130.60897192514588,
      "decode_tokens_per_second_stdev": 1.561483048154041,
      "decode_tokens_per_second_p95": 130.60897192514588,
      "load_seconds_median": 0.531614,
      "load_seconds_mean": 0.5531714400000001,
      "load_seconds_min": 0.501121,
      "load_seconds_max": 0.6309968,
      "load_seconds_stdev": 0.056463033251748,
      "load_seconds_p95": 0.6309968,
      "peak_gpu_memory_used_mib_median": 6011.0,
      "peak_system_memory_used_bytes_median": 43544125440.0,
      "gpu_energy_j_observed_median": 1206.9939522027971
    },
    {
      "system_id": "3090",
      "model": "gemma3:4b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 9.735210300015751,
      "wall_seconds_mean": 9.723726780014113,
      "wall_seconds_min": 9.436986299988348,
      "wall_seconds_max": 10.059978800010867,
      "wall_seconds_stdev": 0.23082308746376654,
      "wall_seconds_p95": 10.059978800010867,
      "ttft_seconds_median": 0.8585128999548033,
      "ttft_seconds_mean": 0.871244759997353,
      "ttft_seconds_min": 0.8408531000022776,
      "ttft_seconds_max": 0.925029500038363,
      "ttft_seconds_stdev": 0.03461446546372492,
      "ttft_seconds_p95": 0.925029500038363,
      "prompt_tokens_median": 32566.0,
      "prompt_tokens_mean": 32566.0,
      "prompt_tokens_min": 32566.0,
      "prompt_tokens_max": 32566.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32566.0,
      "output_tokens_median": 996.0,
      "output_tokens_mean": 996.0,
      "output_tokens_min": 996.0,
      "output_tokens_max": 996.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 996.0,
      "prefill_tokens_per_second_median": 376672.79688168684,
      "prefill_tokens_per_second_mean": 370374.7071991663,
      "prefill_tokens_per_second_min": 312800.76072653226,
      "prefill_tokens_per_second_max": 409243.993163768,
      "prefill_tokens_per_second_stdev": 41211.84454975377,
      "prefill_tokens_per_second_p95": 409243.993163768,
      "decode_tokens_per_second_median": 112.07650188982812,
      "decode_tokens_per_second_mean": 112.5814248192731,
      "decode_tokens_per_second_min": 109.04375632530423,
      "decode_tokens_per_second_max": 116.12749773224917,
      "decode_tokens_per_second_stdev": 2.593294456756201,
      "decode_tokens_per_second_p95": 116.12749773224917,
      "load_seconds_median": 0.5333629,
      "load_seconds_mean": 0.53646642,
      "load_seconds_min": 0.5116779,
      "load_seconds_max": 0.5761791,
      "load_seconds_stdev": 0.023927566598737137,
      "load_seconds_p95": 0.5761791,
      "peak_gpu_memory_used_mib_median": 6985.0,
      "peak_system_memory_used_bytes_median": 43859914752.0,
      "gpu_energy_j_observed_median": 3399.8806562244895
    },
    {
      "system_id": "3090",
      "model": "gemma3:4b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 9.778850500006229,
      "wall_seconds_mean": 9.961344019987155,
      "wall_seconds_min": 9.717304499994498,
      "wall_seconds_max": 10.518512499984354,
      "wall_seconds_stdev": 0.3389817625227423,
      "wall_seconds_p95": 10.518512499984354,
      "ttft_seconds_median": 0.8985363999963738,
      "ttft_seconds_mean": 0.9561262600007467,
      "ttft_seconds_min": 0.8683201000094414,
      "ttft_seconds_max": 1.1779649999807589,
      "ttft_seconds_stdev": 0.12802219473997173,
      "ttft_seconds_p95": 1.1779649999807589,
      "prompt_tokens_median": 32557.0,
      "prompt_tokens_mean": 32557.0,
      "prompt_tokens_min": 32557.0,
      "prompt_tokens_max": 32557.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32557.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 374373.303895865,
      "prefill_tokens_per_second_mean": 345042.79485971684,
      "prefill_tokens_per_second_min": 191534.29815272384,
      "prefill_tokens_per_second_max": 427790.5525261152,
      "prefill_tokens_per_second_stdev": 95750.10768820797,
      "prefill_tokens_per_second_p95": 427790.5525261152,
      "decode_tokens_per_second_median": 112.48523068921051,
      "decode_tokens_per_second_mean": 111.16575586754406,
      "decode_tokens_per_second_min": 107.07348881831555,
      "decode_tokens_per_second_max": 113.21994582199152,
      "decode_tokens_per_second_stdev": 2.645961047618812,
      "decode_tokens_per_second_p95": 113.21994582199152,
      "load_seconds_median": 0.5343985,
      "load_seconds_mean": 0.54927388,
      "load_seconds_min": 0.5301991,
      "load_seconds_max": 0.5784814,
      "load_seconds_stdev": 0.02282202296208644,
      "load_seconds_p95": 0.5784814,
      "peak_gpu_memory_used_mib_median": 6991.0,
      "peak_system_memory_used_bytes_median": 43840729088.0,
      "gpu_energy_j_observed_median": 3565.318722809553
    },
    {
      "system_id": "3090",
      "model": "qwen2.5-coder:14b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 12.782995299960021,
      "wall_seconds_mean": 12.688661879987922,
      "wall_seconds_min": 12.481156199995894,
      "wall_seconds_max": 12.799847699992824,
      "wall_seconds_stdev": 0.14765473109079266,
      "wall_seconds_p95": 12.799847699992824,
      "prompt_tokens_median": 15128.0,
      "prompt_tokens_mean": 15128.0,
      "prompt_tokens_min": 15128.0,
      "prompt_tokens_max": 15128.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15128.0,
      "output_tokens_median": 712.0,
      "output_tokens_mean": 712.0,
      "output_tokens_min": 712.0,
      "output_tokens_max": 712.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 712.0,
      "prefill_tokens_per_second_median": 84863.85208288922,
      "prefill_tokens_per_second_mean": 84385.94548732758,
      "prefill_tokens_per_second_min": 80401.36907672357,
      "prefill_tokens_per_second_max": 86218.58988607154,
      "prefill_tokens_per_second_stdev": 2302.4095954238023,
      "prefill_tokens_per_second_p95": 86218.58988607154,
      "decode_tokens_per_second_median": 63.661230694977675,
      "decode_tokens_per_second_mean": 63.69070894456256,
      "decode_tokens_per_second_min": 63.302843649088864,
      "decode_tokens_per_second_max": 64.18996768503929,
      "decode_tokens_per_second_stdev": 0.36780303590523905,
      "decode_tokens_per_second_p95": 64.18996768503929,
      "prefill_seconds_median": 0.178262,
      "prefill_seconds_mean": 0.1793818,
      "prefill_seconds_min": 0.175461,
      "prefill_seconds_max": 0.188156,
      "prefill_seconds_stdev": 0.0050511242016010585,
      "prefill_seconds_p95": 0.188156,
      "decode_seconds_median": 11.184201,
      "decode_seconds_mean": 11.1793224,
      "decode_seconds_min": 11.092076,
      "decode_seconds_max": 11.24752,
      "decode_seconds_stdev": 0.06446832330920323,
      "decode_seconds_p95": 11.24752,
      "peak_gpu_memory_used_mib_median": 12089.0,
      "peak_system_memory_used_bytes_median": 42171920384.0,
      "gpu_energy_j_observed_median": 4932.851884120702
    },
    {
      "system_id": "3090",
      "model": "qwen2.5-coder:14b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 10.204446599993389,
      "wall_seconds_mean": 10.189163119986187,
      "wall_seconds_min": 10.077290999994148,
      "wall_seconds_max": 10.234034899971448,
      "wall_seconds_stdev": 0.0644263205397144,
      "wall_seconds_p95": 10.234034899971448,
      "ttft_seconds_median": 0.383273699961137,
      "ttft_seconds_mean": 0.3798545399797149,
      "ttft_seconds_min": 0.3394653999712318,
      "ttft_seconds_max": 0.40963329997612163,
      "ttft_seconds_stdev": 0.025581612651698053,
      "ttft_seconds_p95": 0.40963329997612163,
      "prompt_tokens_median": 16025.0,
      "prompt_tokens_mean": 16025.0,
      "prompt_tokens_min": 16025.0,
      "prompt_tokens_max": 16025.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16025.0,
      "output_tokens_median": 524.0,
      "output_tokens_mean": 524.0,
      "output_tokens_min": 524.0,
      "output_tokens_max": 524.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 524.0,
      "prefill_tokens_per_second_median": 643806.998513519,
      "prefill_tokens_per_second_mean": 647207.3896666799,
      "prefill_tokens_per_second_min": 630905.5118110237,
      "prefill_tokens_per_second_max": 675989.2010461487,
      "prefill_tokens_per_second_stdev": 17384.62942964946,
      "prefill_tokens_per_second_p95": 675989.2010461487,
      "decode_tokens_per_second_median": 53.361064793571096,
      "decode_tokens_per_second_mean": 53.42647070335523,
      "decode_tokens_per_second_min": 53.23387673878705,
      "decode_tokens_per_second_max": 53.81815134239313,
      "decode_tokens_per_second_stdev": 0.22567986273808385,
      "decode_tokens_per_second_p95": 53.81815134239313,
      "load_seconds_median": 0.250517,
      "load_seconds_mean": 0.2450655,
      "load_seconds_min": 0.2158411,
      "load_seconds_max": 0.2617597,
      "load_seconds_stdev": 0.01772738210467072,
      "load_seconds_p95": 0.2617597,
      "peak_gpu_memory_used_mib_median": 14417.0,
      "peak_system_memory_used_bytes_median": 42629914624.0,
      "gpu_energy_j_observed_median": 4003.624134478569
    },
    {
      "system_id": "3090",
      "model": "qwen2.5-coder:14b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 42.79090299998643,
      "wall_seconds_mean": 42.96890778000234,
      "wall_seconds_min": 42.39404189999914,
      "wall_seconds_max": 43.93687079998199,
      "wall_seconds_stdev": 0.5832464298389797,
      "wall_seconds_p95": 43.93687079998199,
      "ttft_seconds_median": 22.27993570000399,
      "ttft_seconds_mean": 22.57221766001312,
      "ttft_seconds_min": 22.023156100010965,
      "ttft_seconds_max": 23.724085199995898,
      "ttft_seconds_stdev": 0.6841084984488183,
      "ttft_seconds_p95": 23.724085199995898,
      "prompt_tokens_median": 31994.0,
      "prompt_tokens_mean": 31994.0,
      "prompt_tokens_min": 31994.0,
      "prompt_tokens_max": 31994.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31994.0,
      "output_tokens_median": 898.0,
      "output_tokens_mean": 898.0,
      "output_tokens_min": 898.0,
      "output_tokens_max": 898.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 898.0,
      "prefill_tokens_per_second_median": 1472.5310412525932,
      "prefill_tokens_per_second_mean": 1471.7053147082481,
      "prefill_tokens_per_second_min": 1444.8085091336475,
      "prefill_tokens_per_second_max": 1500.8759488077176,
      "prefill_tokens_per_second_stdev": 21.11394184977594,
      "prefill_tokens_per_second_p95": 1500.8759488077176,
      "decode_tokens_per_second_median": 44.08864547505172,
      "decode_tokens_per_second_mean": 44.03414254278862,
      "decode_tokens_per_second_min": 43.72229835755579,
      "decode_tokens_per_second_max": 44.43228356703165,
      "decode_tokens_per_second_stdev": 0.28680018152017434,
      "decode_tokens_per_second_p95": 44.43228356703165,
      "load_seconds_median": 0.2379115,
      "load_seconds_mean": 0.23647602,
      "load_seconds_min": 0.1941602,
      "load_seconds_max": 0.2885598,
      "load_seconds_stdev": 0.036176956877714286,
      "load_seconds_p95": 0.2885598,
      "peak_gpu_memory_used_mib_median": 16744.0,
      "peak_system_memory_used_bytes_median": 48660422656.0,
      "gpu_energy_j_observed_median": 17488.425433937304
    },
    {
      "system_id": "3090",
      "model": "qwen2.5vl:7b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 9.099808000028133,
      "wall_seconds_mean": 9.01255778002087,
      "wall_seconds_min": 8.746828800009098,
      "wall_seconds_max": 9.166103199997451,
      "wall_seconds_stdev": 0.1763821852818659,
      "wall_seconds_p95": 9.166103199997451,
      "prompt_tokens_median": 15294.0,
      "prompt_tokens_mean": 15294.0,
      "prompt_tokens_min": 15294.0,
      "prompt_tokens_max": 15294.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15294.0,
      "output_tokens_median": 870.0,
      "output_tokens_mean": 870.0,
      "output_tokens_min": 870.0,
      "output_tokens_max": 870.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 870.0,
      "prefill_tokens_per_second_median": 163336.35926736798,
      "prefill_tokens_per_second_mean": 163405.93271781577,
      "prefill_tokens_per_second_min": 159058.9997191975,
      "prefill_tokens_per_second_max": 167697.3684210526,
      "prefill_tokens_per_second_stdev": 3401.3161587501236,
      "prefill_tokens_per_second_p95": 167697.3684210526,
      "decode_tokens_per_second_median": 114.86020390986775,
      "decode_tokens_per_second_mean": 115.21406782333717,
      "decode_tokens_per_second_min": 113.22589020602557,
      "decode_tokens_per_second_max": 118.03463380365386,
      "decode_tokens_per_second_stdev": 1.801814165286888,
      "decode_tokens_per_second_p95": 118.03463380365386,
      "prefill_seconds_median": 0.093635,
      "prefill_seconds_mean": 0.0936276,
      "prefill_seconds_min": 0.0912,
      "prefill_seconds_max": 0.096153,
      "prefill_seconds_stdev": 0.0019499690510364526,
      "prefill_seconds_p95": 0.096153,
      "decode_seconds_median": 7.574425,
      "decode_seconds_mean": 7.5526256,
      "decode_seconds_min": 7.370718,
      "decode_seconds_max": 7.683755,
      "decode_seconds_stdev": 0.11705794991712418,
      "decode_seconds_p95": 7.683755,
      "peak_gpu_memory_used_mib_median": 14124.0,
      "peak_system_memory_used_bytes_median": 44384051200.0,
      "gpu_energy_j_observed_median": 3342.1711628532407
    },
    {
      "system_id": "3090",
      "model": "qwen2.5vl:7b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 7.261640699987765,
      "wall_seconds_mean": 7.323952119995374,
      "wall_seconds_min": 7.16484969999874,
      "wall_seconds_max": 7.497420199972112,
      "wall_seconds_stdev": 0.14392147851019096,
      "wall_seconds_p95": 7.497420199972112,
      "ttft_seconds_median": 0.36617250001290813,
      "ttft_seconds_mean": 0.3753004599944688,
      "ttft_seconds_min": 0.35619039996527135,
      "ttft_seconds_max": 0.41525799996452406,
      "ttft_seconds_stdev": 0.02416293745513602,
      "ttft_seconds_p95": 0.41525799996452406,
      "prompt_tokens_median": 16015.0,
      "prompt_tokens_mean": 16015.0,
      "prompt_tokens_min": 16015.0,
      "prompt_tokens_max": 16015.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16015.0,
      "output_tokens_median": 721.0,
      "output_tokens_mean": 721.0,
      "output_tokens_min": 721.0,
      "output_tokens_max": 721.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 721.0,
      "prefill_tokens_per_second_median": 1041219.6866263571,
      "prefill_tokens_per_second_mean": 1027788.8045231969,
      "prefill_tokens_per_second_min": 930509.5578409156,
      "prefill_tokens_per_second_max": 1076204.5561454203,
      "prefill_tokens_per_second_stdev": 57118.268290959815,
      "prefill_tokens_per_second_p95": 1076204.5561454203,
      "decode_tokens_per_second_median": 104.58198175747917,
      "decode_tokens_per_second_mean": 103.80685456181861,
      "decode_tokens_per_second_min": 101.81729035523361,
      "decode_tokens_per_second_max": 105.9139966595225,
      "decode_tokens_per_second_stdev": 1.829738785674161,
      "decode_tokens_per_second_p95": 105.9139966595225,
      "load_seconds_median": 0.2574908,
      "load_seconds_mean": 0.26452184,
      "load_seconds_min": 0.2499076,
      "load_seconds_max": 0.296476,
      "load_seconds_stdev": 0.019019395381241748,
      "load_seconds_p95": 0.296476,
      "peak_gpu_memory_used_mib_median": 14818.0,
      "peak_system_memory_used_bytes_median": 43690700800.0,
      "gpu_energy_j_observed_median": 2917.9206222140792
    },
    {
      "system_id": "3090",
      "model": "qwen2.5vl:7b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 2.161503800016362,
      "wall_seconds_mean": 2.1737187599996104,
      "wall_seconds_min": 2.1514431999530643,
      "wall_seconds_max": 2.2361019999952987,
      "wall_seconds_stdev": 0.03532893746625229,
      "wall_seconds_p95": 2.2361019999952987,
      "ttft_seconds_median": 0.2878678999841213,
      "ttft_seconds_mean": 0.293762920005247,
      "ttft_seconds_min": 0.27243620000081137,
      "ttft_seconds_max": 0.3177817999967374,
      "ttft_seconds_stdev": 0.02009019633530335,
      "ttft_seconds_p95": 0.3177817999967374,
      "prompt_tokens_median": 4053.0,
      "prompt_tokens_mean": 4053.0,
      "prompt_tokens_min": 4053.0,
      "prompt_tokens_max": 4053.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4053.0,
      "output_tokens_median": 219.0,
      "output_tokens_mean": 219.0,
      "output_tokens_min": 219.0,
      "output_tokens_max": 219.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 219.0,
      "prefill_tokens_per_second_median": 307488.05098247476,
      "prefill_tokens_per_second_mean": 310131.93189813127,
      "prefill_tokens_per_second_min": 287548.7761617595,
      "prefill_tokens_per_second_max": 338285.61889658624,
      "prefill_tokens_per_second_stdev": 18314.33285982466,
      "prefill_tokens_per_second_p95": 338285.61889658624,
      "decode_tokens_per_second_median": 116.4188292305885,
      "decode_tokens_per_second_mean": 116.54502761923055,
      "decode_tokens_per_second_min": 114.25655762208626,
      "decode_tokens_per_second_max": 118.18187705199355,
      "decode_tokens_per_second_stdev": 1.496021434087526,
      "decode_tokens_per_second_p95": 118.18187705199355,
      "load_seconds_median": 0.2447413,
      "load_seconds_mean": 0.24761088,
      "load_seconds_min": 0.2256763,
      "load_seconds_max": 0.2713958,
      "load_seconds_stdev": 0.018477330535875584,
      "load_seconds_p95": 0.2713958,
      "peak_gpu_memory_used_mib_median": 14124.0,
      "peak_system_memory_used_bytes_median": 44026568704.0,
      "gpu_energy_j_observed_median": 812.7290371143818
    },
    {
      "system_id": "3090",
      "model": "qwen2.5vl:7b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 11.248735799978022,
      "wall_seconds_mean": 11.303928299993277,
      "wall_seconds_min": 11.055825699993875,
      "wall_seconds_max": 11.643633100029547,
      "wall_seconds_stdev": 0.22481269015599115,
      "wall_seconds_p95": 11.643633100029547,
      "ttft_seconds_median": 0.4853267999715172,
      "ttft_seconds_mean": 0.4791473199846223,
      "ttft_seconds_min": 0.4420090999919921,
      "ttft_seconds_max": 0.5067179999896325,
      "ttft_seconds_stdev": 0.029040261912085263,
      "ttft_seconds_p95": 0.5067179999896325,
      "prompt_tokens_median": 31984.0,
      "prompt_tokens_mean": 31984.0,
      "prompt_tokens_min": 31984.0,
      "prompt_tokens_max": 31984.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31984.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 1609986.9123124937,
      "prefill_tokens_per_second_mean": 1648747.1537854697,
      "prefill_tokens_per_second_min": 1527849.4315467658,
      "prefill_tokens_per_second_max": 1814900.981671679,
      "prefill_tokens_per_second_stdev": 108254.1985816149,
      "prefill_tokens_per_second_p95": 1814900.981671679,
      "decode_tokens_per_second_median": 93.11090132190478,
      "decode_tokens_per_second_mean": 92.42517303476095,
      "decode_tokens_per_second_min": 89.79733459959547,
      "decode_tokens_per_second_max": 94.35821833189578,
      "decode_tokens_per_second_stdev": 1.7392557478365744,
      "decode_tokens_per_second_p95": 94.35821833189578,
      "load_seconds_median": 0.2737853,
      "load_seconds_mean": 0.26563148000000003,
      "load_seconds_min": 0.2277083,
      "load_seconds_max": 0.295556,
      "load_seconds_stdev": 0.02933514521383864,
      "load_seconds_p95": 0.295556,
      "peak_gpu_memory_used_mib_median": 16107.0,
      "peak_system_memory_used_bytes_median": 44565454848.0,
      "gpu_energy_j_observed_median": 4480.7980636978145
    },
    {
      "system_id": "3090",
      "model": "qwen2.5vl:7b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 11.436872599995695,
      "wall_seconds_mean": 11.462223920004908,
      "wall_seconds_min": 11.299816099985037,
      "wall_seconds_max": 11.67697520001093,
      "wall_seconds_stdev": 0.15832371912257887,
      "wall_seconds_p95": 11.67697520001093,
      "ttft_seconds_median": 0.4886201999615878,
      "ttft_seconds_mean": 0.49173821998992934,
      "ttft_seconds_min": 0.4795643999823369,
      "ttft_seconds_max": 0.503941900038626,
      "ttft_seconds_stdev": 0.01024902971603981,
      "ttft_seconds_p95": 0.503941900038626,
      "prompt_tokens_median": 31979.0,
      "prompt_tokens_mean": 31979.0,
      "prompt_tokens_min": 31979.0,
      "prompt_tokens_max": 31979.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31979.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 1687902.4596220837,
      "prefill_tokens_per_second_mean": 1676996.733068923,
      "prefill_tokens_per_second_min": 1605210.320248971,
      "prefill_tokens_per_second_max": 1762608.1684396185,
      "prefill_tokens_per_second_stdev": 62045.771082716434,
      "prefill_tokens_per_second_p95": 1762608.1684396185,
      "decode_tokens_per_second_median": 91.35178164293255,
      "decode_tokens_per_second_mean": 91.18847128926902,
      "decode_tokens_per_second_min": 89.52901738411789,
      "decode_tokens_per_second_max": 92.49152846967863,
      "decode_tokens_per_second_stdev": 1.2952044966253584,
      "decode_tokens_per_second_p95": 92.49152846967863,
      "load_seconds_median": 0.2755871,
      "load_seconds_mean": 0.27583112,
      "load_seconds_min": 0.2650886,
      "load_seconds_max": 0.2888119,
      "load_seconds_stdev": 0.00877140570558676,
      "load_seconds_p95": 0.2888119,
      "peak_gpu_memory_used_mib_median": 16113.0,
      "peak_system_memory_used_bytes_median": 43836182528.0,
      "gpu_energy_j_observed_median": 4517.877874860764
    },
    {
      "system_id": "spark",
      "model": "aero-specialist-verification-v5-e1:v4",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 38.77896889997646,
      "wall_seconds_mean": 39.44669732002076,
      "wall_seconds_min": 38.53873270004988,
      "wall_seconds_max": 42.33554040000308,
      "wall_seconds_stdev": 1.6191531271147173,
      "wall_seconds_p95": 42.33554040000308,
      "prompt_tokens_median": 15314.0,
      "prompt_tokens_mean": 15314.0,
      "prompt_tokens_min": 15314.0,
      "prompt_tokens_max": 15314.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15314.0,
      "output_tokens_median": 870.0,
      "output_tokens_mean": 870.0,
      "output_tokens_min": 870.0,
      "output_tokens_max": 870.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 870.0,
      "prefill_tokens_per_second_median": 32250.855909685353,
      "prefill_tokens_per_second_mean": 26515.256013258353,
      "prefill_tokens_per_second_min": 3685.131272384357,
      "prefill_tokens_per_second_max": 32370.0781709909,
      "prefill_tokens_per_second_stdev": 12763.100228851557,
      "prefill_tokens_per_second_p95": 32370.0781709909,
      "decode_tokens_per_second_median": 24.62174149117339,
      "decode_tokens_per_second_mean": 24.619277560851764,
      "decode_tokens_per_second_min": 24.6068997142657,
      "decode_tokens_per_second_max": 24.63123045992774,
      "decode_tokens_per_second_stdev": 0.009665181733478283,
      "decode_tokens_per_second_p95": 24.63123045992774,
      "prefill_seconds_median": 0.474840111,
      "prefill_seconds_mean": 1.2113330094,
      "prefill_seconds_min": 0.473091227,
      "prefill_seconds_max": 4.155618584,
      "prefill_seconds_stdev": 1.645906813388097,
      "prefill_seconds_p95": 4.155618584,
      "decode_seconds_median": 35.334624901,
      "decode_seconds_mean": 35.338165595,
      "decode_seconds_min": 35.321012542,
      "decode_seconds_max": 35.355937159999996,
      "decode_seconds_stdev": 0.013873853413742724,
      "decode_seconds_p95": 35.355937159999996,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 21638909952.0,
      "gpu_energy_j_observed_median": 1470.3592932188512
    },
    {
      "system_id": "spark",
      "model": "aero-specialist-verification-v5-e1:v4",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 31.112300399981905,
      "wall_seconds_mean": 31.110283920005894,
      "wall_seconds_min": 31.085892000002787,
      "wall_seconds_max": 31.131041000015102,
      "wall_seconds_stdev": 0.018423811528793674,
      "wall_seconds_p95": 31.131041000015102,
      "ttft_seconds_median": 0.17277790000662208,
      "ttft_seconds_mean": 0.17517926002619788,
      "ttft_seconds_min": 0.15903219999745488,
      "ttft_seconds_max": 0.18924220005283132,
      "ttft_seconds_stdev": 0.011564231183364224,
      "ttft_seconds_p95": 0.18924220005283132,
      "prompt_tokens_median": 16009.0,
      "prompt_tokens_mean": 16009.0,
      "prompt_tokens_min": 16009.0,
      "prompt_tokens_max": 16009.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16009.0,
      "output_tokens_median": 624.0,
      "output_tokens_mean": 624.0,
      "output_tokens_min": 624.0,
      "output_tokens_max": 624.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 624.0,
      "prefill_tokens_per_second_median": 305132.0876571736,
      "prefill_tokens_per_second_mean": 305787.193525049,
      "prefill_tokens_per_second_min": 303335.42586586595,
      "prefill_tokens_per_second_max": 310905.32118273363,
      "prefill_tokens_per_second_stdev": 3115.5360474117556,
      "prefill_tokens_per_second_p95": 310905.32118273363,
      "decode_tokens_per_second_median": 20.281017890141896,
      "decode_tokens_per_second_mean": 20.280179632499962,
      "decode_tokens_per_second_min": 20.269299843959995,
      "decode_tokens_per_second_max": 20.28774625774089,
      "decode_tokens_per_second_stdev": 0.007878421566572359,
      "decode_tokens_per_second_p95": 20.28774625774089,
      "load_seconds_median": 0.080504161,
      "load_seconds_mean": 0.081710283,
      "load_seconds_min": 0.070270352,
      "load_seconds_max": 0.090440477,
      "load_seconds_stdev": 0.008639080947804316,
      "load_seconds_p95": 0.090440477,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 23360479232.0,
      "gpu_energy_j_observed_median": 1372.469312888384
    },
    {
      "system_id": "spark",
      "model": "aero-specialist-verification-v5-e1:v4",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 32.687195499951486,
      "wall_seconds_mean": 32.67796999999555,
      "wall_seconds_min": 32.59242960001575,
      "wall_seconds_max": 32.72138790000463,
      "wall_seconds_stdev": 0.05176755260003535,
      "wall_seconds_p95": 32.72138790000463,
      "ttft_seconds_median": 0.20115329994587228,
      "ttft_seconds_mean": 0.20833454000530766,
      "ttft_seconds_min": 0.17945730005158111,
      "ttft_seconds_max": 0.23408999998355284,
      "ttft_seconds_stdev": 0.02289021808040723,
      "ttft_seconds_p95": 0.23408999998355284,
      "prompt_tokens_median": 31978.0,
      "prompt_tokens_mean": 31978.0,
      "prompt_tokens_min": 31978.0,
      "prompt_tokens_max": 31978.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31978.0,
      "output_tokens_median": 543.0,
      "output_tokens_mean": 543.0,
      "output_tokens_min": 543.0,
      "output_tokens_max": 543.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 543.0,
      "prefill_tokens_per_second_median": 508523.1132644353,
      "prefill_tokens_per_second_mean": 507934.46728637867,
      "prefill_tokens_per_second_min": 500089.71839121927,
      "prefill_tokens_per_second_max": 513691.811225954,
      "prefill_tokens_per_second_stdev": 5105.130897157127,
      "prefill_tokens_per_second_p95": 513691.811225954,
      "decode_tokens_per_second_median": 16.81821765258613,
      "decode_tokens_per_second_mean": 16.81194745213141,
      "decode_tokens_per_second_min": 16.789668882691075,
      "decode_tokens_per_second_max": 16.841253244943136,
      "decode_tokens_per_second_stdev": 0.022093048506048454,
      "decode_tokens_per_second_p95": 16.841253244943136,
      "load_seconds_median": 0.076744099,
      "load_seconds_mean": 0.07759026699999999,
      "load_seconds_min": 0.054377589,
      "load_seconds_max": 0.093785092,
      "load_seconds_stdev": 0.01549692377613633,
      "load_seconds_p95": 0.093785092,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 25833684992.0,
      "gpu_energy_j_observed_median": 1470.340517948866
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 27.396292500023264,
      "wall_seconds_mean": 27.198941780021414,
      "wall_seconds_min": 26.820001400017645,
      "wall_seconds_max": 27.49878359999275,
      "wall_seconds_stdev": 0.3426617489945916,
      "wall_seconds_p95": 27.49878359999275,
      "prompt_tokens_median": 15146.0,
      "prompt_tokens_mean": 15146.0,
      "prompt_tokens_min": 15146.0,
      "prompt_tokens_max": 15146.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15146.0,
      "output_tokens_median": 525.0,
      "output_tokens_mean": 525.0,
      "output_tokens_min": 525.0,
      "output_tokens_max": 525.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 525.0,
      "prefill_tokens_per_second_median": 7435.590149051519,
      "prefill_tokens_per_second_mean": 7440.744439943265,
      "prefill_tokens_per_second_min": 7414.449646031648,
      "prefill_tokens_per_second_max": 7471.648798787843,
      "prefill_tokens_per_second_stdev": 23.938004039061564,
      "prefill_tokens_per_second_p95": 7471.648798787843,
      "decode_tokens_per_second_median": 24.19809605126013,
      "decode_tokens_per_second_mean": 24.189645758542106,
      "decode_tokens_per_second_min": 24.157347423060237,
      "decode_tokens_per_second_max": 24.200025502733617,
      "decode_tokens_per_second_stdev": 0.018176262923067586,
      "decode_tokens_per_second_p95": 24.200025502733617,
      "prefill_seconds_median": 2.0369600390000002,
      "prefill_seconds_mean": 2.035565856,
      "prefill_seconds_min": 2.027129541,
      "prefill_seconds_max": 2.042767936,
      "prefill_seconds_stdev": 0.006544269695718283,
      "prefill_seconds_p95": 2.042767936,
      "decode_seconds_median": 21.695921815,
      "decode_seconds_mean": 21.703510775199998,
      "decode_seconds_min": 21.694192014000002,
      "decode_seconds_max": 21.732518509,
      "decode_seconds_stdev": 0.01632409220964235,
      "decode_seconds_p95": 21.732518509,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 21992476672.0,
      "gpu_energy_j_observed_median": 1192.2831065511702
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 4096,
      "n": 5,
      "wall_seconds_median": 5.501201900013257,
      "wall_seconds_mean": 5.503906579990871,
      "wall_seconds_min": 5.4736230999697,
      "wall_seconds_max": 5.530512199969962,
      "wall_seconds_stdev": 0.02430873145696648,
      "wall_seconds_p95": 5.530512199969962,
      "ttft_seconds_median": 2.1274371999897994,
      "ttft_seconds_mean": 2.1286626799963413,
      "ttft_seconds_min": 2.11580969998613,
      "ttft_seconds_max": 2.1390023000421934,
      "ttft_seconds_stdev": 0.008776446403003155,
      "ttft_seconds_p95": 2.1390023000421934,
      "prompt_tokens_median": 3720.0,
      "prompt_tokens_mean": 3720.0,
      "prompt_tokens_min": 3720.0,
      "prompt_tokens_max": 3720.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 3720.0,
      "output_tokens_median": 81.0,
      "output_tokens_mean": 81.0,
      "output_tokens_min": 81.0,
      "output_tokens_max": 81.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 81.0,
      "prefill_tokens_per_second_median": 1919.3568386305706,
      "prefill_tokens_per_second_mean": 1918.7146897118175,
      "prefill_tokens_per_second_min": 1913.8964996043612,
      "prefill_tokens_per_second_max": 1922.621713761549,
      "prefill_tokens_per_second_stdev": 3.4257156084086793,
      "prefill_tokens_per_second_p95": 1922.621713761549,
      "decode_tokens_per_second_median": 24.395816388794895,
      "decode_tokens_per_second_mean": 24.408948198991133,
      "decode_tokens_per_second_min": 24.298893532873755,
      "decode_tokens_per_second_max": 24.481310514799876,
      "decode_tokens_per_second_stdev": 0.07396761275492152,
      "decode_tokens_per_second_p95": 24.481310514799876,
      "load_seconds_median": 0.16036853,
      "load_seconds_mean": 0.1620756664,
      "load_seconds_min": 0.148883932,
      "load_seconds_max": 0.17416505,
      "load_seconds_stdev": 0.010124096980409895,
      "load_seconds_p95": 0.17416505,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 21726654464.0,
      "gpu_energy_j_observed_median": 328.7156189405918
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 7.613619200012181,
      "wall_seconds_mean": 7.601256619987543,
      "wall_seconds_min": 7.553142299992032,
      "wall_seconds_max": 7.6377936999779195,
      "wall_seconds_stdev": 0.032133344702326254,
      "wall_seconds_p95": 7.6377936999779195,
      "ttft_seconds_median": 4.310302399971988,
      "ttft_seconds_mean": 4.316628480004146,
      "ttft_seconds_min": 4.2832163000130095,
      "ttft_seconds_max": 4.365589399996679,
      "ttft_seconds_stdev": 0.03052427959859738,
      "ttft_seconds_p95": 4.365589399996679,
      "prompt_tokens_median": 7790.0,
      "prompt_tokens_mean": 7790.0,
      "prompt_tokens_min": 7790.0,
      "prompt_tokens_max": 7790.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 7790.0,
      "output_tokens_median": 76.0,
      "output_tokens_mean": 76.0,
      "output_tokens_min": 76.0,
      "output_tokens_max": 76.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 76.0,
      "prefill_tokens_per_second_median": 1900.9601304516052,
      "prefill_tokens_per_second_mean": 1902.2648665604695,
      "prefill_tokens_per_second_min": 1899.194959512978,
      "prefill_tokens_per_second_max": 1907.8012566589896,
      "prefill_tokens_per_second_stdev": 3.437262897557336,
      "prefill_tokens_per_second_p95": 1907.8012566589896,
      "decode_tokens_per_second_median": 23.5726838997051,
      "decode_tokens_per_second_mean": 23.551287508861904,
      "decode_tokens_per_second_min": 23.4547708631419,
      "decode_tokens_per_second_max": 23.585791430395766,
      "decode_tokens_per_second_stdev": 0.05456170512704877,
      "decode_tokens_per_second_p95": 23.585791430395766,
      "load_seconds_median": 0.16389136,
      "load_seconds_mean": 0.1667443274,
      "load_seconds_min": 0.130910392,
      "load_seconds_max": 0.199659226,
      "load_seconds_stdev": 0.02491517996999514,
      "load_seconds_p95": 0.199659226,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 22175645696.0,
      "gpu_energy_j_observed_median": 488.0224811184406
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 16384,
      "n": 5,
      "wall_seconds_median": 12.098093700013123,
      "wall_seconds_mean": 12.087600460008252,
      "wall_seconds_min": 12.040089700021781,
      "wall_seconds_max": 12.132126800017431,
      "wall_seconds_stdev": 0.038310662783105354,
      "wall_seconds_p95": 12.132126800017431,
      "ttft_seconds_median": 8.742501400003675,
      "ttft_seconds_mean": 8.749498259997926,
      "ttft_seconds_min": 8.724438600009307,
      "ttft_seconds_max": 8.78070200001821,
      "ttft_seconds_stdev": 0.025351500288718666,
      "ttft_seconds_p95": 8.78070200001821,
      "prompt_tokens_median": 15930.0,
      "prompt_tokens_mean": 15930.0,
      "prompt_tokens_min": 15930.0,
      "prompt_tokens_max": 15930.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15930.0,
      "output_tokens_median": 73.0,
      "output_tokens_mean": 73.0,
      "output_tokens_min": 73.0,
      "output_tokens_max": 73.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 73.0,
      "prefill_tokens_per_second_median": 1878.4045326015987,
      "prefill_tokens_per_second_mean": 1878.7598153741599,
      "prefill_tokens_per_second_min": 1875.5388041518806,
      "prefill_tokens_per_second_max": 1881.9502471596509,
      "prefill_tokens_per_second_stdev": 2.62486840760317,
      "prefill_tokens_per_second_p95": 1881.9502471596509,
      "decode_tokens_per_second_median": 22.225691087655736,
      "decode_tokens_per_second_mean": 22.24454006993666,
      "decode_tokens_per_second_min": 22.197441849550113,
      "decode_tokens_per_second_max": 22.31964069340984,
      "decode_tokens_per_second_stdev": 0.05251728816652165,
      "decode_tokens_per_second_p95": 22.31964069340984,
      "load_seconds_median": 0.164777008,
      "load_seconds_mean": 0.1675186698,
      "load_seconds_min": 0.144740956,
      "load_seconds_max": 0.197781229,
      "load_seconds_stdev": 0.01907273347574161,
      "load_seconds_p95": 0.197781229,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 23536414720.0,
      "gpu_energy_j_observed_median": 887.2105590569973
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 32768,
      "n": 5,
      "wall_seconds_median": 21.952726599993184,
      "wall_seconds_mean": 21.94743185999105,
      "wall_seconds_min": 21.90614999999525,
      "wall_seconds_max": 22.00619809998898,
      "wall_seconds_stdev": 0.04153023067364845,
      "wall_seconds_p95": 22.00619809998898,
      "ttft_seconds_median": 18.403323899954557,
      "ttft_seconds_mean": 18.390337640000507,
      "ttft_seconds_min": 18.348710099991877,
      "ttft_seconds_max": 18.414860300021246,
      "ttft_seconds_stdev": 0.02796699020744371,
      "ttft_seconds_p95": 18.414860300021246,
      "prompt_tokens_median": 32210.0,
      "prompt_tokens_mean": 32210.0,
      "prompt_tokens_min": 32210.0,
      "prompt_tokens_max": 32210.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32210.0,
      "output_tokens_median": 70.0,
      "output_tokens_mean": 70.0,
      "output_tokens_min": 70.0,
      "output_tokens_max": 70.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 70.0,
      "prefill_tokens_per_second_median": 1790.5728077211606,
      "prefill_tokens_per_second_mean": 1790.6030137865796,
      "prefill_tokens_per_second_min": 1789.7333407145618,
      "prefill_tokens_per_second_max": 1791.8910316342922,
      "prefill_tokens_per_second_stdev": 0.854973434427421,
      "prefill_tokens_per_second_p95": 1791.8910316342922,
      "decode_tokens_per_second_median": 19.9936265631109,
      "decode_tokens_per_second_mean": 19.987282437180546,
      "decode_tokens_per_second_min": 19.954953562397108,
      "decode_tokens_per_second_max": 20.014025188403227,
      "decode_tokens_per_second_stdev": 0.024463617141623054,
      "decode_tokens_per_second_p95": 20.014025188403227,
      "load_seconds_median": 0.172744958,
      "load_seconds_mean": 0.168901149,
      "load_seconds_min": 0.141376223,
      "load_seconds_max": 0.185953533,
      "load_seconds_stdev": 0.016475323936135995,
      "load_seconds_p95": 0.185953533,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 25203544064.0,
      "gpu_energy_j_observed_median": 1754.6665540921688
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 65536,
      "n": 5,
      "wall_seconds_median": 44.87468130001798,
      "wall_seconds_mean": 44.860289800004104,
      "wall_seconds_min": 44.77661299996544,
      "wall_seconds_max": 44.938310500001535,
      "wall_seconds_stdev": 0.07265927836848143,
      "wall_seconds_p95": 44.938310500001535,
      "ttft_seconds_median": 40.26792770001339,
      "ttft_seconds_mean": 40.24517777999863,
      "ttft_seconds_min": 40.172325299994554,
      "ttft_seconds_max": 40.31653109996114,
      "ttft_seconds_stdev": 0.061624687730487696,
      "ttft_seconds_p95": 40.31653109996114,
      "prompt_tokens_median": 64715.0,
      "prompt_tokens_mean": 64715.0,
      "prompt_tokens_min": 64715.0,
      "prompt_tokens_max": 64715.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 64715.0,
      "output_tokens_median": 77.0,
      "output_tokens_mean": 77.0,
      "output_tokens_min": 77.0,
      "output_tokens_max": 77.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 77.0,
      "prefill_tokens_per_second_median": 1636.017214902574,
      "prefill_tokens_per_second_mean": 1636.8224412615998,
      "prefill_tokens_per_second_min": 1635.4117295540211,
      "prefill_tokens_per_second_max": 1640.199611550118,
      "prefill_tokens_per_second_stdev": 1.9513327191150807,
      "prefill_tokens_per_second_p95": 1640.199611550118,
      "decode_tokens_per_second_median": 16.90517772461752,
      "decode_tokens_per_second_mean": 16.906912687577815,
      "decode_tokens_per_second_min": 16.899079086835954,
      "decode_tokens_per_second_max": 16.91829891287256,
      "decode_tokens_per_second_stdev": 0.007156307860222009,
      "decode_tokens_per_second_p95": 16.91829891287256,
      "load_seconds_median": 0.187169328,
      "load_seconds_mean": 0.1818337052,
      "load_seconds_min": 0.167913369,
      "load_seconds_max": 0.197259397,
      "load_seconds_stdev": 0.01280733509583608,
      "load_seconds_p95": 0.197259397,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 28239949824.0,
      "gpu_energy_j_observed_median": 3755.4869084715842
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 131072,
      "n": 5,
      "wall_seconds_median": 99.04780160001246,
      "wall_seconds_mean": 99.00619157999753,
      "wall_seconds_min": 98.89559109997936,
      "wall_seconds_max": 99.06824649998453,
      "wall_seconds_stdev": 0.07262266127817416,
      "wall_seconds_p95": 99.06824649998453,
      "ttft_seconds_median": 93.23713520000456,
      "ttft_seconds_mean": 93.22840816000243,
      "ttft_seconds_min": 93.12931360001676,
      "ttft_seconds_max": 93.29855780000798,
      "ttft_seconds_stdev": 0.06694068743731753,
      "ttft_seconds_p95": 93.29855780000798,
      "prompt_tokens_median": 129780.0,
      "prompt_tokens_mean": 129780.0,
      "prompt_tokens_min": 129780.0,
      "prompt_tokens_max": 129780.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 129780.0,
      "output_tokens_median": 74.0,
      "output_tokens_mean": 74.0,
      "output_tokens_min": 74.0,
      "output_tokens_max": 74.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 74.0,
      "prefill_tokens_per_second_median": 1413.2171628023918,
      "prefill_tokens_per_second_mean": 1413.0596370279693,
      "prefill_tokens_per_second_min": 1412.1403395906993,
      "prefill_tokens_per_second_max": 1413.6129217507048,
      "prefill_tokens_per_second_stdev": 0.6201312723185632,
      "prefill_tokens_per_second_p95": 1413.6129217507048,
      "decode_tokens_per_second_median": 12.919148516205023,
      "decode_tokens_per_second_mean": 12.92006739456108,
      "decode_tokens_per_second_min": 12.914077583532183,
      "decode_tokens_per_second_max": 12.926322078653415,
      "decode_tokens_per_second_stdev": 0.004517283258763278,
      "decode_tokens_per_second_p95": 12.926322078653415,
      "load_seconds_median": 0.181139881,
      "load_seconds_mean": 0.1768539292,
      "load_seconds_min": 0.146755365,
      "load_seconds_max": 0.199714033,
      "load_seconds_stdev": 0.019891039647664106,
      "load_seconds_p95": 0.199714033,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 35129073664.0,
      "gpu_energy_j_observed_median": 8284.923661469222
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 54.696462899970356,
      "wall_seconds_mean": 54.702057299984155,
      "wall_seconds_min": 54.648396599979606,
      "wall_seconds_max": 54.764334100007545,
      "wall_seconds_stdev": 0.041403562854132846,
      "wall_seconds_p95": 54.764334100007545,
      "ttft_seconds_median": 9.114731399982702,
      "ttft_seconds_mean": 9.11453153998591,
      "ttft_seconds_min": 9.078029899974354,
      "ttft_seconds_max": 9.171149100002367,
      "ttft_seconds_stdev": 0.03719337592335195,
      "ttft_seconds_p95": 9.171149100002367,
      "prompt_tokens_median": 16308.0,
      "prompt_tokens_mean": 16308.0,
      "prompt_tokens_min": 16308.0,
      "prompt_tokens_max": 16308.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16308.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 1852.2480435337006,
      "prefill_tokens_per_second_mean": 1851.2777119508792,
      "prefill_tokens_per_second_min": 1847.0198583245005,
      "prefill_tokens_per_second_max": 1854.4921713716606,
      "prefill_tokens_per_second_stdev": 3.0826075182414567,
      "prefill_tokens_per_second_p95": 1854.4921713716606,
      "decode_tokens_per_second_median": 22.14313873296103,
      "decode_tokens_per_second_mean": 22.139935473391983,
      "decode_tokens_per_second_min": 22.129707950823587,
      "decode_tokens_per_second_max": 22.145178465254336,
      "decode_tokens_per_second_stdev": 0.006298889555993124,
      "decode_tokens_per_second_p95": 22.145178465254336,
      "load_seconds_median": 0.187220382,
      "load_seconds_mean": 0.1833007556,
      "load_seconds_min": 0.15201197,
      "load_seconds_max": 0.208945459,
      "load_seconds_stdev": 0.02171077704282952,
      "load_seconds_p95": 0.208945459,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 23515156480.0,
      "gpu_energy_j_observed_median": 2920.858564103842
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 6.960013400006574,
      "wall_seconds_mean": 6.9547031400026755,
      "wall_seconds_min": 6.9356514000101015,
      "wall_seconds_max": 6.96378689998528,
      "wall_seconds_stdev": 0.011491988773486601,
      "wall_seconds_p95": 6.96378689998528,
      "ttft_seconds_median": 2.348633499990683,
      "ttft_seconds_mean": 2.3501651800004764,
      "ttft_seconds_min": 2.3181875000009313,
      "ttft_seconds_max": 2.384216899983585,
      "ttft_seconds_stdev": 0.023509515249025642,
      "ttft_seconds_p95": 2.384216899983585,
      "prompt_tokens_median": 4112.0,
      "prompt_tokens_mean": 4112.0,
      "prompt_tokens_min": 4112.0,
      "prompt_tokens_max": 4112.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4112.0,
      "output_tokens_median": 115.0,
      "output_tokens_mean": 115.0,
      "output_tokens_min": 115.0,
      "output_tokens_max": 115.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 115.0,
      "prefill_tokens_per_second_median": 1901.5614151775949,
      "prefill_tokens_per_second_mean": 1902.2715145559873,
      "prefill_tokens_per_second_min": 1898.240082760904,
      "prefill_tokens_per_second_max": 1906.9712194310646,
      "prefill_tokens_per_second_stdev": 3.5344956925612347,
      "prefill_tokens_per_second_p95": 1906.9712194310646,
      "decode_tokens_per_second_median": 25.38691963032847,
      "decode_tokens_per_second_mean": 25.379261183061292,
      "decode_tokens_per_second_min": 25.340824280779263,
      "decode_tokens_per_second_max": 25.42049412593003,
      "decode_tokens_per_second_stdev": 0.030098957428339408,
      "decode_tokens_per_second_p95": 25.42049412593003,
      "load_seconds_median": 0.164101518,
      "load_seconds_mean": 0.1637178736,
      "load_seconds_min": 0.130332092,
      "load_seconds_max": 0.193998743,
      "load_seconds_stdev": 0.02270931238126789,
      "load_seconds_p95": 0.193998743,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 48377782272.0,
      "gpu_energy_j_observed_median": 391.99785560369486
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 68.26462600001832,
      "wall_seconds_mean": 68.26868746000109,
      "wall_seconds_min": 68.1673305000295,
      "wall_seconds_max": 68.35391910001636,
      "wall_seconds_stdev": 0.06761275498887735,
      "wall_seconds_p95": 68.35391910001636,
      "ttft_seconds_median": 18.568253099976573,
      "ttft_seconds_mean": 18.58423553999746,
      "ttft_seconds_min": 18.52992510003969,
      "ttft_seconds_max": 18.634360099968035,
      "ttft_seconds_stdev": 0.04329069623076725,
      "ttft_seconds_p95": 18.634360099968035,
      "prompt_tokens_median": 32565.0,
      "prompt_tokens_mean": 32565.0,
      "prompt_tokens_min": 32565.0,
      "prompt_tokens_max": 32565.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32565.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 1789.8269545147848,
      "prefill_tokens_per_second_mean": 1789.4502175698135,
      "prefill_tokens_per_second_min": 1788.662150984916,
      "prefill_tokens_per_second_max": 1790.2273529217891,
      "prefill_tokens_per_second_stdev": 0.7270450326924504,
      "prefill_tokens_per_second_p95": 1790.2273529217891,
      "decode_tokens_per_second_median": 20.324941577084996,
      "decode_tokens_per_second_mean": 20.31158943663835,
      "decode_tokens_per_second_min": 20.261477535272086,
      "decode_tokens_per_second_max": 20.33204086654584,
      "decode_tokens_per_second_stdev": 0.028829964886661962,
      "decode_tokens_per_second_p95": 20.33204086654584,
      "load_seconds_median": 0.158522801,
      "load_seconds_mean": 0.1671504086,
      "load_seconds_min": 0.129011841,
      "load_seconds_max": 0.205212952,
      "load_seconds_stdev": 0.02971307046592241,
      "load_seconds_p95": 0.205212952,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 25448681472.0,
      "gpu_energy_j_observed_median": 4027.035607172251
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 69.77884599997196,
      "wall_seconds_mean": 69.778426139988,
      "wall_seconds_min": 69.64366330002667,
      "wall_seconds_max": 69.89773229998536,
      "wall_seconds_stdev": 0.09265241954100263,
      "wall_seconds_p95": 69.89773229998536,
      "ttft_seconds_median": 18.667896599974483,
      "ttft_seconds_mean": 18.660802659997717,
      "ttft_seconds_min": 18.575277500029188,
      "ttft_seconds_max": 18.72083589999238,
      "ttft_seconds_stdev": 0.05769613759480938,
      "ttft_seconds_p95": 18.72083589999238,
      "prompt_tokens_median": 32556.0,
      "prompt_tokens_mean": 32556.0,
      "prompt_tokens_min": 32556.0,
      "prompt_tokens_max": 32556.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32556.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 1783.3871949882305,
      "prefill_tokens_per_second_mean": 1784.5961803418177,
      "prefill_tokens_per_second_min": 1779.6017604753379,
      "prefill_tokens_per_second_max": 1792.1205564989878,
      "prefill_tokens_per_second_stdev": 5.108653686135364,
      "prefill_tokens_per_second_p95": 1792.1205564989878,
      "decode_tokens_per_second_median": 19.72602021668627,
      "decode_tokens_per_second_mean": 19.73161279531152,
      "decode_tokens_per_second_min": 19.721572637297353,
      "decode_tokens_per_second_max": 19.75763583815387,
      "decode_tokens_per_second_stdev": 0.014715391335030942,
      "decode_tokens_per_second_p95": 19.75763583815387,
      "load_seconds_median": 0.172553383,
      "load_seconds_mean": 0.1726917858,
      "load_seconds_min": 0.143279841,
      "load_seconds_max": 0.192895002,
      "load_seconds_stdev": 0.02106958047297582,
      "load_seconds_p95": 0.192895002,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 25423945728.0,
      "gpu_energy_j_observed_median": 3983.250004781484
    },
    {
      "system_id": "spark",
      "model": "gemma3:4b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 13.403588699991815,
      "wall_seconds_mean": 13.437139460002072,
      "wall_seconds_min": 13.347203399986029,
      "wall_seconds_max": 13.653546700021252,
      "wall_seconds_stdev": 0.12442524963488595,
      "wall_seconds_p95": 13.653546700021252,
      "prompt_tokens_median": 15309.0,
      "prompt_tokens_mean": 15309.0,
      "prompt_tokens_min": 15309.0,
      "prompt_tokens_max": 15309.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15309.0,
      "output_tokens_median": 643.0,
      "output_tokens_mean": 643.0,
      "output_tokens_min": 643.0,
      "output_tokens_max": 643.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 643.0,
      "prefill_tokens_per_second_median": 21916.059008773424,
      "prefill_tokens_per_second_mean": 21940.607399967845,
      "prefill_tokens_per_second_min": 21848.601199075038,
      "prefill_tokens_per_second_max": 22102.242246849528,
      "prefill_tokens_per_second_stdev": 101.27455691851907,
      "prefill_tokens_per_second_p95": 22102.242246849528,
      "decode_tokens_per_second_median": 67.27403380941357,
      "decode_tokens_per_second_mean": 67.31432462370181,
      "decode_tokens_per_second_min": 67.16245591924337,
      "decode_tokens_per_second_max": 67.5158436941866,
      "decode_tokens_per_second_stdev": 0.16071028856307834,
      "decode_tokens_per_second_p95": 67.5158436941866,
      "prefill_seconds_median": 0.6985288730000001,
      "prefill_seconds_mean": 0.6977591722,
      "prefill_seconds_min": 0.692644657,
      "prefill_seconds_max": 0.700685589,
      "prefill_seconds_stdev": 0.0032098186694219572,
      "prefill_seconds_p95": 0.700685589,
      "decode_seconds_median": 9.557922479,
      "decode_seconds_mean": 9.5522451418,
      "decode_seconds_min": 9.523690512,
      "decode_seconds_max": 9.573801184,
      "decode_seconds_stdev": 0.022792273052317426,
      "decode_seconds_p95": 9.573801184,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 16059179008.0,
      "gpu_energy_j_observed_median": 429.06734197020535
    },
    {
      "system_id": "spark",
      "model": "gemma3:4b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 15.039582900004461,
      "wall_seconds_mean": 15.048877700010781,
      "wall_seconds_min": 15.02641940000467,
      "wall_seconds_max": 15.080306800024118,
      "wall_seconds_stdev": 0.022621201585608937,
      "wall_seconds_p95": 15.080306800024118,
      "ttft_seconds_median": 3.1641445999848656,
      "ttft_seconds_mean": 3.1687761000008323,
      "ttft_seconds_min": 3.1461836000089534,
      "ttft_seconds_max": 3.1948142999899574,
      "ttft_seconds_stdev": 0.021082669752264905,
      "ttft_seconds_p95": 3.1948142999899574,
      "prompt_tokens_median": 16308.0,
      "prompt_tokens_mean": 16308.0,
      "prompt_tokens_min": 16308.0,
      "prompt_tokens_max": 16308.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16308.0,
      "output_tokens_median": 729.0,
      "output_tokens_mean": 729.0,
      "output_tokens_min": 729.0,
      "output_tokens_max": 729.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 729.0,
      "prefill_tokens_per_second_median": 5696.816168703371,
      "prefill_tokens_per_second_mean": 5694.173846782061,
      "prefill_tokens_per_second_min": 5670.24064041275,
      "prefill_tokens_per_second_max": 5705.039484288611,
      "prefill_tokens_per_second_stdev": 14.217146570745015,
      "prefill_tokens_per_second_p95": 5705.039484288611,
      "decode_tokens_per_second_median": 63.053190594232014,
      "decode_tokens_per_second_mean": 63.049471906537285,
      "decode_tokens_per_second_min": 62.97133273224389,
      "decode_tokens_per_second_max": 63.1599314989424,
      "decode_tokens_per_second_stdev": 0.07495092574829858,
      "decode_tokens_per_second_p95": 63.1599314989424,
      "load_seconds_median": 0.173340694,
      "load_seconds_mean": 0.1751594668,
      "load_seconds_min": 0.156661941,
      "load_seconds_max": 0.200234841,
      "load_seconds_stdev": 0.017177552672466317,
      "load_seconds_p95": 0.200234841,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 17069637632.0,
      "gpu_energy_j_observed_median": 702.1070633029939
    },
    {
      "system_id": "spark",
      "model": "gemma3:4b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 5.8957928000018,
      "wall_seconds_mean": 5.906690039997921,
      "wall_seconds_min": 5.884041299985256,
      "wall_seconds_max": 5.932959500001743,
      "wall_seconds_stdev": 0.022435822878166558,
      "wall_seconds_p95": 5.932959500001743,
      "ttft_seconds_median": 0.9116524999844842,
      "ttft_seconds_mean": 0.9209780999924988,
      "ttft_seconds_min": 0.9083134999964386,
      "ttft_seconds_max": 0.9394960000063293,
      "ttft_seconds_stdev": 0.015393988044943099,
      "ttft_seconds_p95": 0.9394960000063293,
      "prompt_tokens_median": 4112.0,
      "prompt_tokens_mean": 4112.0,
      "prompt_tokens_min": 4112.0,
      "prompt_tokens_max": 4112.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4112.0,
      "output_tokens_median": 342.0,
      "output_tokens_mean": 342.0,
      "output_tokens_min": 342.0,
      "output_tokens_max": 342.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 342.0,
      "prefill_tokens_per_second_median": 5783.4522264321295,
      "prefill_tokens_per_second_mean": 5784.655262776332,
      "prefill_tokens_per_second_min": 5772.615507172452,
      "prefill_tokens_per_second_max": 5801.293115007273,
      "prefill_tokens_per_second_stdev": 11.95369001937274,
      "prefill_tokens_per_second_p95": 5801.293115007273,
      "decode_tokens_per_second_median": 71.1289251438206,
      "decode_tokens_per_second_mean": 71.16291952662473,
      "decode_tokens_per_second_min": 71.08197094304083,
      "decode_tokens_per_second_max": 71.26999576242567,
      "decode_tokens_per_second_stdev": 0.07482851376493523,
      "decode_tokens_per_second_p95": 71.26999576242567,
      "load_seconds_median": 0.176041667,
      "load_seconds_mean": 0.1826403554,
      "load_seconds_min": 0.158704253,
      "load_seconds_max": 0.204869977,
      "load_seconds_stdev": 0.01928261670140965,
      "load_seconds_p95": 0.204869977,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 38195003392.0,
      "gpu_energy_j_observed_median": 241.97279193520544
    },
    {
      "system_id": "spark",
      "model": "gemma3:4b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 23.388107399980072,
      "wall_seconds_mean": 23.390238399989904,
      "wall_seconds_min": 23.344776499958243,
      "wall_seconds_max": 23.43585880001774,
      "wall_seconds_stdev": 0.03760674679960412,
      "wall_seconds_p95": 23.43585880001774,
      "ttft_seconds_median": 6.299955900001805,
      "ttft_seconds_mean": 6.295649800007231,
      "ttft_seconds_min": 6.2683367999852635,
      "ttft_seconds_max": 6.324865300033707,
      "ttft_seconds_stdev": 0.022641534123767845,
      "ttft_seconds_p95": 6.324865300033707,
      "prompt_tokens_median": 32565.0,
      "prompt_tokens_mean": 32565.0,
      "prompt_tokens_min": 32565.0,
      "prompt_tokens_max": 32565.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32565.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 5527.812924481268,
      "prefill_tokens_per_second_mean": 5529.668952016216,
      "prefill_tokens_per_second_min": 5523.513969236686,
      "prefill_tokens_per_second_max": 5540.174306973242,
      "prefill_tokens_per_second_stdev": 6.247843891952216,
      "prefill_tokens_per_second_p95": 5540.174306973242,
      "decode_tokens_per_second_median": 60.13543214208414,
      "decode_tokens_per_second_mean": 60.14189177504434,
      "decode_tokens_per_second_min": 60.082151895632705,
      "decode_tokens_per_second_max": 60.25100015488375,
      "decode_tokens_per_second_stdev": 0.0675965792807018,
      "decode_tokens_per_second_p95": 60.25100015488375,
      "load_seconds_median": 0.175931397,
      "load_seconds_mean": 0.172825568,
      "load_seconds_min": 0.157692977,
      "load_seconds_max": 0.183919676,
      "load_seconds_stdev": 0.011670634851335869,
      "load_seconds_p95": 0.183919676,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 40165699584.0,
      "gpu_energy_j_observed_median": 1199.2185351574424
    },
    {
      "system_id": "spark",
      "model": "gemma3:4b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 23.56643940001959,
      "wall_seconds_mean": 23.568563219998033,
      "wall_seconds_min": 23.476335599960294,
      "wall_seconds_max": 23.644589799980167,
      "wall_seconds_stdev": 0.06541828562359413,
      "wall_seconds_p95": 23.644589799980167,
      "ttft_seconds_median": 6.323581500037108,
      "ttft_seconds_mean": 6.3270984999951905,
      "ttft_seconds_min": 6.284043399966322,
      "ttft_seconds_max": 6.3994794999598525,
      "ttft_seconds_stdev": 0.04575270232319688,
      "ttft_seconds_p95": 6.3994794999598525,
      "prompt_tokens_median": 32556.0,
      "prompt_tokens_mean": 32556.0,
      "prompt_tokens_min": 32556.0,
      "prompt_tokens_max": 32556.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 32556.0,
      "output_tokens_median": 959.0,
      "output_tokens_mean": 959.0,
      "output_tokens_min": 959.0,
      "output_tokens_max": 959.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 959.0,
      "prefill_tokens_per_second_median": 5467.367446158801,
      "prefill_tokens_per_second_mean": 5472.051952792185,
      "prefill_tokens_per_second_min": 5460.863685540756,
      "prefill_tokens_per_second_max": 5484.58503105641,
      "prefill_tokens_per_second_stdev": 9.912901160697784,
      "prefill_tokens_per_second_p95": 5484.58503105641,
      "decode_tokens_per_second_median": 57.086654193132325,
      "decode_tokens_per_second_mean": 57.09358289248874,
      "decode_tokens_per_second_min": 57.00723810719411,
      "decode_tokens_per_second_max": 57.18132378343071,
      "decode_tokens_per_second_stdev": 0.07414952086064626,
      "decode_tokens_per_second_p95": 57.18132378343071,
      "load_seconds_median": 0.158136956,
      "load_seconds_mean": 0.1646635372,
      "load_seconds_min": 0.142523585,
      "load_seconds_max": 0.197139812,
      "load_seconds_stdev": 0.020902574716028972,
      "load_seconds_p95": 0.197139812,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 18167279616.0,
      "gpu_energy_j_observed_median": 1161.0176533305646
    },
    {
      "system_id": "spark",
      "model": "qwen2.5vl:7b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 24.57027799997013,
      "wall_seconds_mean": 24.57302913998719,
      "wall_seconds_min": 24.479171499959193,
      "wall_seconds_max": 24.706208299961872,
      "wall_seconds_stdev": 0.08408826505326963,
      "wall_seconds_p95": 24.706208299961872,
      "prompt_tokens_median": 15294.0,
      "prompt_tokens_mean": 15294.0,
      "prompt_tokens_min": 15294.0,
      "prompt_tokens_max": 15294.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15294.0,
      "output_tokens_median": 870.0,
      "output_tokens_mean": 870.0,
      "output_tokens_min": 870.0,
      "output_tokens_max": 870.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 870.0,
      "prefill_tokens_per_second_median": 76757.67047847748,
      "prefill_tokens_per_second_mean": 76706.18521231637,
      "prefill_tokens_per_second_min": 76342.4881956663,
      "prefill_tokens_per_second_max": 76877.40295606432,
      "prefill_tokens_per_second_stdev": 215.12218334206935,
      "prefill_tokens_per_second_p95": 76877.40295606432,
      "decode_tokens_per_second_median": 40.15977939178132,
      "decode_tokens_per_second_mean": 40.16457221010843,
      "decode_tokens_per_second_min": 40.087932995298836,
      "decode_tokens_per_second_max": 40.29138304183784,
      "decode_tokens_per_second_stdev": 0.08195423711468446,
      "decode_tokens_per_second_p95": 40.29138304183784,
      "prefill_seconds_median": 0.199250445,
      "prefill_seconds_mean": 0.1993854402,
      "prefill_seconds_min": 0.198940123,
      "prefill_seconds_max": 0.200334052,
      "prefill_seconds_stdev": 0.0005607555566026122,
      "prefill_seconds_p95": 0.200334052,
      "decode_seconds_median": 21.663465616,
      "decode_seconds_mean": 21.6609525866,
      "decode_seconds_min": 21.592706289000002,
      "decode_seconds_max": 21.702291313,
      "decode_seconds_stdev": 0.044145181671854,
      "decode_seconds_p95": 21.702291313,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 27191349248.0,
      "gpu_energy_j_observed_median": 1066.741272279024
    },
    {
      "system_id": "spark",
      "model": "qwen2.5vl:7b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 24.151167000003625,
      "wall_seconds_mean": 24.154775900009554,
      "wall_seconds_min": 24.12690550001571,
      "wall_seconds_max": 24.18344450002769,
      "wall_seconds_stdev": 0.02062894497365567,
      "wall_seconds_p95": 24.18344450002769,
      "ttft_seconds_median": 0.17950859997654334,
      "ttft_seconds_mean": 0.17152141999686138,
      "ttft_seconds_min": 0.14929020003182814,
      "ttft_seconds_max": 0.1899834999931045,
      "ttft_seconds_stdev": 0.016830157656121822,
      "ttft_seconds_p95": 0.1899834999931045,
      "prompt_tokens_median": 16015.0,
      "prompt_tokens_mean": 16015.0,
      "prompt_tokens_min": 16015.0,
      "prompt_tokens_max": 16015.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16015.0,
      "output_tokens_median": 769.0,
      "output_tokens_mean": 769.0,
      "output_tokens_min": 769.0,
      "output_tokens_max": 769.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 769.0,
      "prefill_tokens_per_second_median": 489823.2528686622,
      "prefill_tokens_per_second_mean": 486032.46008690854,
      "prefill_tokens_per_second_min": 469599.77051041875,
      "prefill_tokens_per_second_max": 497741.9330375952,
      "prefill_tokens_per_second_stdev": 10764.70296917355,
      "prefill_tokens_per_second_p95": 497741.9330375952,
      "decode_tokens_per_second_median": 32.374204629230334,
      "decode_tokens_per_second_mean": 32.3727266299309,
      "decode_tokens_per_second_min": 32.36550379700679,
      "decode_tokens_per_second_max": 32.37886782956181,
      "decode_tokens_per_second_stdev": 0.006116479980025489,
      "decode_tokens_per_second_p95": 32.37886782956181,
      "load_seconds_median": 0.102945779,
      "load_seconds_mean": 0.0968954076,
      "load_seconds_min": 0.084375955,
      "load_seconds_max": 0.106030265,
      "load_seconds_stdev": 0.010682427482287288,
      "load_seconds_p95": 0.106030265,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 29825994752.0,
      "gpu_energy_j_observed_median": 1332.9609519422054
    },
    {
      "system_id": "spark",
      "model": "qwen2.5vl:7b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 5.3058400000445545,
      "wall_seconds_mean": 5.312536740000359,
      "wall_seconds_min": 5.295013100025244,
      "wall_seconds_max": 5.332832099986263,
      "wall_seconds_stdev": 0.016658054735564566,
      "wall_seconds_p95": 5.332832099986263,
      "ttft_seconds_median": 0.12267630000133067,
      "ttft_seconds_mean": 0.11449284000555053,
      "ttft_seconds_min": 0.0945260000298731,
      "ttft_seconds_max": 0.1279958999948576,
      "ttft_seconds_stdev": 0.01614021686773989,
      "ttft_seconds_p95": 0.1279958999948576,
      "prompt_tokens_median": 4053.0,
      "prompt_tokens_mean": 4053.0,
      "prompt_tokens_min": 4053.0,
      "prompt_tokens_max": 4053.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 4053.0,
      "output_tokens_median": 210.0,
      "output_tokens_mean": 210.0,
      "output_tokens_min": 210.0,
      "output_tokens_max": 210.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 210.0,
      "prefill_tokens_per_second_median": 156671.16489050607,
      "prefill_tokens_per_second_mean": 156028.79724051311,
      "prefill_tokens_per_second_min": 153554.0720793347,
      "prefill_tokens_per_second_max": 158286.20058716252,
      "prefill_tokens_per_second_stdev": 1962.449723225858,
      "prefill_tokens_per_second_p95": 158286.20058716252,
      "decode_tokens_per_second_median": 41.1292041004281,
      "decode_tokens_per_second_mean": 41.13486457909118,
      "decode_tokens_per_second_min": 41.10580684498221,
      "decode_tokens_per_second_max": 41.1768886709116,
      "decode_tokens_per_second_stdev": 0.027402532933568768,
      "decode_tokens_per_second_p95": 41.1768886709116,
      "load_seconds_median": 0.084983329,
      "load_seconds_mean": 0.0755209462,
      "load_seconds_min": 0.053063014,
      "load_seconds_max": 0.090540795,
      "load_seconds_stdev": 0.017409089010076997,
      "load_seconds_p95": 0.090540795,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 34364604416.0,
      "gpu_energy_j_observed_median": 238.23749910712243
    },
    {
      "system_id": "spark",
      "model": "qwen2.5vl:7b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 39.15960640000412,
      "wall_seconds_mean": 39.17011492000893,
      "wall_seconds_min": 39.09613720001653,
      "wall_seconds_max": 39.28840730001684,
      "wall_seconds_stdev": 0.07140969640689108,
      "wall_seconds_p95": 39.28840730001684,
      "ttft_seconds_median": 0.1811472999979742,
      "ttft_seconds_mean": 0.18232986001530663,
      "ttft_seconds_min": 0.1597015000297688,
      "ttft_seconds_max": 0.19782300002407283,
      "ttft_seconds_stdev": 0.01500677240623016,
      "ttft_seconds_p95": 0.19782300002407283,
      "prompt_tokens_median": 31984.0,
      "prompt_tokens_mean": 31984.0,
      "prompt_tokens_min": 31984.0,
      "prompt_tokens_max": 31984.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31984.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 747434.0547304287,
      "prefill_tokens_per_second_mean": 760665.927886558,
      "prefill_tokens_per_second_min": 740808.0644684604,
      "prefill_tokens_per_second_max": 797678.4723280091,
      "prefill_tokens_per_second_stdev": 23818.053468872455,
      "prefill_tokens_per_second_p95": 797678.4723280091,
      "decode_tokens_per_second_median": 25.85502401197771,
      "decode_tokens_per_second_mean": 25.84569989062763,
      "decode_tokens_per_second_min": 25.78900120272505,
      "decode_tokens_per_second_max": 25.877249498346345,
      "decode_tokens_per_second_stdev": 0.033873346119974,
      "decode_tokens_per_second_p95": 25.877249498346345,
      "load_seconds_median": 0.060180439,
      "load_seconds_mean": 0.0680119292,
      "load_seconds_min": 0.057844244,
      "load_seconds_max": 0.101198976,
      "load_seconds_stdev": 0.01858466998856527,
      "load_seconds_p95": 0.101198976,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 43781812224.0,
      "gpu_energy_j_observed_median": 2387.9249830722815
    },
    {
      "system_id": "spark",
      "model": "qwen2.5vl:7b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 39.98556170001393,
      "wall_seconds_mean": 39.97792683999287,
      "wall_seconds_min": 39.95924629998626,
      "wall_seconds_max": 39.98879339999985,
      "wall_seconds_stdev": 0.013209009476358778,
      "wall_seconds_p95": 39.98879339999985,
      "ttft_seconds_median": 0.20248219999484718,
      "ttft_seconds_mean": 0.20757741999113932,
      "ttft_seconds_min": 0.19708319997880608,
      "ttft_seconds_max": 0.2195740999886766,
      "ttft_seconds_stdev": 0.01083971696636673,
      "ttft_seconds_p95": 0.2195740999886766,
      "prompt_tokens_median": 31979.0,
      "prompt_tokens_mean": 31979.0,
      "prompt_tokens_min": 31979.0,
      "prompt_tokens_max": 31979.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31979.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 748465.3990103003,
      "prefill_tokens_per_second_mean": 748704.121131426,
      "prefill_tokens_per_second_min": 748070.4593995688,
      "prefill_tokens_per_second_max": 749411.5365366655,
      "prefill_tokens_per_second_stdev": 604.7471002947648,
      "prefill_tokens_per_second_p95": 749411.5365366655,
      "decode_tokens_per_second_median": 25.33095324331622,
      "decode_tokens_per_second_mean": 25.334622744130286,
      "decode_tokens_per_second_min": 25.328208255428088,
      "decode_tokens_per_second_max": 25.3449498992134,
      "decode_tokens_per_second_stdev": 0.007619750818213886,
      "decode_tokens_per_second_p95": 25.3449498992134,
      "load_seconds_median": 0.081204725,
      "load_seconds_mean": 0.0845625884,
      "load_seconds_min": 0.063515136,
      "load_seconds_max": 0.108827023,
      "load_seconds_stdev": 0.017857618634440447,
      "load_seconds_p95": 0.108827023,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 36968321024.0,
      "gpu_energy_j_observed_median": 2378.959251116514
    },
    {
      "system_id": "spark",
      "model": "qwen3-coder:30b-a3b-q4_K_M",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 15.026726400014013,
      "wall_seconds_mean": 15.032175960007589,
      "wall_seconds_min": 14.995993999997154,
      "wall_seconds_max": 15.101637799991295,
      "wall_seconds_stdev": 0.04233964933458347,
      "wall_seconds_p95": 15.101637799991295,
      "prompt_tokens_median": 15294.0,
      "prompt_tokens_mean": 15294.0,
      "prompt_tokens_min": 15294.0,
      "prompt_tokens_max": 15294.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 15294.0,
      "output_tokens_median": 870.0,
      "output_tokens_mean": 870.0,
      "output_tokens_min": 870.0,
      "output_tokens_max": 870.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 870.0,
      "prefill_tokens_per_second_median": 55216.99254530526,
      "prefill_tokens_per_second_mean": 55137.18938729583,
      "prefill_tokens_per_second_min": 54681.620341476955,
      "prefill_tokens_per_second_max": 55318.69443844531,
      "prefill_tokens_per_second_stdev": 258.12021916419735,
      "prefill_tokens_per_second_p95": 55318.69443844531,
      "decode_tokens_per_second_median": 71.61741137806472,
      "decode_tokens_per_second_mean": 71.58615685862844,
      "decode_tokens_per_second_min": 71.43312686217227,
      "decode_tokens_per_second_max": 71.71441028939961,
      "decode_tokens_per_second_stdev": 0.1355881393820036,
      "decode_tokens_per_second_p95": 71.71441028939961,
      "prefill_seconds_median": 0.276979953,
      "prefill_seconds_mean": 0.2773857336,
      "prefill_seconds_min": 0.276470733,
      "prefill_seconds_max": 0.279691785,
      "prefill_seconds_stdev": 0.0013062094007760862,
      "prefill_seconds_p95": 0.279691785,
      "decode_seconds_median": 12.147883919,
      "decode_seconds_mean": 12.1532225766,
      "decode_seconds_min": 12.13145303,
      "decode_seconds_max": 12.179223257,
      "decode_seconds_stdev": 0.023026600001215404,
      "decode_seconds_p95": 12.179223257,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 31412097024.0,
      "gpu_energy_j_observed_median": 670.9777383220196
    },
    {
      "system_id": "spark",
      "model": "qwen3-coder:30b-a3b-q4_K_M",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 21.53753059997689,
      "wall_seconds_mean": 21.535121739993336,
      "wall_seconds_min": 21.49870729999384,
      "wall_seconds_max": 21.58058930002153,
      "wall_seconds_stdev": 0.03148606336189136,
      "wall_seconds_p95": 21.58058930002153,
      "ttft_seconds_median": 0.15814569999929518,
      "ttft_seconds_mean": 0.14821505999425427,
      "ttft_seconds_min": 0.12010189995635301,
      "ttft_seconds_max": 0.1652468999964185,
      "ttft_seconds_stdev": 0.018626382536964418,
      "ttft_seconds_p95": 0.1652468999964185,
      "prompt_tokens_median": 16004.0,
      "prompt_tokens_mean": 16004.0,
      "prompt_tokens_min": 16004.0,
      "prompt_tokens_max": 16004.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 16004.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 610605.8466005805,
      "prefill_tokens_per_second_mean": 606642.9034257189,
      "prefill_tokens_per_second_min": 571427.3268444696,
      "prefill_tokens_per_second_max": 625660.0774412677,
      "prefill_tokens_per_second_stdev": 21389.719030417953,
      "prefill_tokens_per_second_p95": 625660.0774412677,
      "decode_tokens_per_second_median": 47.40268316864143,
      "decode_tokens_per_second_mean": 47.40158048267745,
      "decode_tokens_per_second_min": 47.361635109210646,
      "decode_tokens_per_second_max": 47.44624495169178,
      "decode_tokens_per_second_stdev": 0.03142808785285281,
      "decode_tokens_per_second_p95": 47.44624495169178,
      "load_seconds_median": 0.074388568,
      "load_seconds_mean": 0.07337405079999999,
      "load_seconds_min": 0.053787283,
      "load_seconds_max": 0.098231261,
      "load_seconds_stdev": 0.018860030938986916,
      "load_seconds_p95": 0.098231261,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 32696320000.0,
      "gpu_energy_j_observed_median": 780.2369675934316
    },
    {
      "system_id": "spark",
      "model": "qwen3-coder:30b-a3b-q4_K_M",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 29.45893680001609,
      "wall_seconds_mean": 29.458411900012287,
      "wall_seconds_min": 29.438212500012014,
      "wall_seconds_max": 29.48841570003424,
      "wall_seconds_stdev": 0.019198378982888176,
      "wall_seconds_p95": 29.48841570003424,
      "ttft_seconds_median": 0.19226260000141338,
      "ttft_seconds_mean": 0.19470222000963985,
      "ttft_seconds_min": 0.18062820000341162,
      "ttft_seconds_max": 0.21975860005477443,
      "ttft_seconds_stdev": 0.014967432515909623,
      "ttft_seconds_p95": 0.21975860005477443,
      "prompt_tokens_median": 31973.0,
      "prompt_tokens_mean": 31973.0,
      "prompt_tokens_min": 31973.0,
      "prompt_tokens_max": 31973.0,
      "prompt_tokens_stdev": 0.0,
      "prompt_tokens_p95": 31973.0,
      "output_tokens_median": 1000.0,
      "output_tokens_mean": 1000.0,
      "output_tokens_min": 1000.0,
      "output_tokens_max": 1000.0,
      "output_tokens_stdev": 0.0,
      "output_tokens_p95": 1000.0,
      "prefill_tokens_per_second_median": 921500.7749592966,
      "prefill_tokens_per_second_mean": 926161.6543559779,
      "prefill_tokens_per_second_min": 897640.1102424349,
      "prefill_tokens_per_second_max": 970221.2506826491,
      "prefill_tokens_per_second_stdev": 29008.30692901473,
      "prefill_tokens_per_second_p95": 970221.2506826491,
      "decode_tokens_per_second_median": 34.47154932750504,
      "decode_tokens_per_second_mean": 34.4778525681765,
      "decode_tokens_per_second_min": 34.452630002468275,
      "decode_tokens_per_second_max": 34.518959374713134,
      "decode_tokens_per_second_stdev": 0.025748654986126194,
      "decode_tokens_per_second_p95": 34.518959374713134,
      "load_seconds_median": 0.076964346,
      "load_seconds_mean": 0.0803426282,
      "load_seconds_min": 0.068247405,
      "load_seconds_max": 0.098883641,
      "load_seconds_stdev": 0.012925756916131357,
      "load_seconds_p95": 0.098883641,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 34379694080.0,
      "gpu_energy_j_observed_median": 995.3633492875099
    }
  ]
}
