{
  "schema": "aero.public.inference_benchmark.v1",
  "suite": "cache-neutral",
  "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": 15.250062100007199,
      "wall_seconds_mean": 15.310228959994856,
      "wall_seconds_min": 13.294200700009242,
      "wall_seconds_max": 17.01875409996137,
      "wall_seconds_stdev": 1.5587532673654698,
      "wall_seconds_p95": 17.01875409996137,
      "prompt_tokens_median": 15368.0,
      "prompt_tokens_mean": 15383.0,
      "prompt_tokens_min": 15292.0,
      "prompt_tokens_max": 15458.0,
      "prompt_tokens_stdev": 73.64102117705865,
      "prompt_tokens_p95": 15458.0,
      "output_tokens_median": 555.0,
      "output_tokens_mean": 544.8,
      "output_tokens_min": 428.0,
      "output_tokens_max": 633.0,
      "output_tokens_stdev": 82.31464025311658,
      "output_tokens_p95": 633.0,
      "prefill_tokens_per_second_median": 5455.050145427775,
      "prefill_tokens_per_second_mean": 5355.755767719014,
      "prefill_tokens_per_second_min": 4903.200021569259,
      "prefill_tokens_per_second_max": 5652.367774244306,
      "prefill_tokens_per_second_stdev": 297.2036930334741,
      "prefill_tokens_per_second_p95": 5652.367774244306,
      "decode_tokens_per_second_median": 63.002455623749334,
      "decode_tokens_per_second_mean": 63.105874650084786,
      "decode_tokens_per_second_min": 62.310606865447504,
      "decode_tokens_per_second_max": 64.20948972769676,
      "decode_tokens_per_second_stdev": 0.7004021505583445,
      "decode_tokens_per_second_p95": 64.20948972769676,
      "prefill_seconds_median": 2.817206,
      "prefill_seconds_mean": 2.8797585998,
      "prefill_seconds_min": 2.71373,
      "prefill_seconds_max": 3.152635,
      "prefill_seconds_stdev": 0.17036202815038864,
      "prefill_seconds_p95": 3.152635,
      "decode_seconds_median": 8.781147,
      "decode_seconds_mean": 8.6394523998,
      "decode_seconds_min": 6.793386,
      "decode_seconds_max": 10.079099,
      "decode_seconds_stdev": 1.3559664534960998,
      "decode_seconds_p95": 10.079099,
      "peak_gpu_memory_used_mib_median": 11467.0,
      "peak_system_memory_used_bytes_median": 50810404864.0,
      "gpu_energy_j_observed_median": 4991.832356247902
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 4096,
      "n": 5,
      "wall_seconds_median": 3.8892678000265732,
      "wall_seconds_mean": 3.9239291800069624,
      "wall_seconds_min": 3.794923799985554,
      "wall_seconds_max": 4.066444299998693,
      "wall_seconds_stdev": 0.10910304795343757,
      "wall_seconds_p95": 4.066444299998693,
      "ttft_seconds_median": 2.8154166000313126,
      "ttft_seconds_mean": 2.824477059999481,
      "ttft_seconds_min": 2.699208899983205,
      "ttft_seconds_max": 2.9379825000069104,
      "ttft_seconds_stdev": 0.09256091384044365,
      "ttft_seconds_p95": 2.9379825000069104,
      "prompt_tokens_median": 3755.0,
      "prompt_tokens_mean": 3755.0,
      "prompt_tokens_min": 3754.0,
      "prompt_tokens_max": 3756.0,
      "prompt_tokens_stdev": 0.7071067811865476,
      "prompt_tokens_p95": 3756.0,
      "output_tokens_median": 72.0,
      "output_tokens_mean": 72.0,
      "output_tokens_min": 70.0,
      "output_tokens_max": 73.0,
      "output_tokens_stdev": 1.224744871391589,
      "output_tokens_p95": 73.0,
      "prefill_tokens_per_second_median": 2473.3367057658056,
      "prefill_tokens_per_second_mean": 2457.885501607207,
      "prefill_tokens_per_second_min": 2348.1534719081656,
      "prefill_tokens_per_second_max": 2519.826437307658,
      "prefill_tokens_per_second_stdev": 69.08176667964976,
      "prefill_tokens_per_second_p95": 2519.826437307658,
      "decode_tokens_per_second_median": 65.06663003656384,
      "decode_tokens_per_second_mean": 65.50541346893469,
      "decode_tokens_per_second_min": 64.52110108843513,
      "decode_tokens_per_second_max": 66.64268747135733,
      "decode_tokens_per_second_stdev": 1.0435036433228404,
      "decode_tokens_per_second_p95": 66.64268747135733,
      "load_seconds_median": 0.5834082,
      "load_seconds_mean": 0.5734107,
      "load_seconds_min": 0.5321438,
      "load_seconds_max": 0.5929104,
      "load_seconds_stdev": 0.024352730406465743,
      "load_seconds_p95": 0.5929104,
      "peak_gpu_memory_used_mib_median": 10902.0,
      "peak_system_memory_used_bytes_median": 49595109376.0,
      "gpu_energy_j_observed_median": 1255.7060613453389
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 5.413977899996098,
      "wall_seconds_mean": 5.380829600000288,
      "wall_seconds_min": 5.228105399990454,
      "wall_seconds_max": 5.521683400031179,
      "wall_seconds_stdev": 0.122154328061734,
      "wall_seconds_p95": 5.521683400031179,
      "ttft_seconds_median": 4.28658740001265,
      "ttft_seconds_mean": 4.252587740006857,
      "ttft_seconds_min": 4.119694299995899,
      "ttft_seconds_max": 4.389525400009006,
      "ttft_seconds_stdev": 0.11549995394728844,
      "ttft_seconds_p95": 4.389525400009006,
      "prompt_tokens_median": 7825.0,
      "prompt_tokens_mean": 7824.6,
      "prompt_tokens_min": 7823.0,
      "prompt_tokens_max": 7826.0,
      "prompt_tokens_stdev": 1.140175425099138,
      "prompt_tokens_p95": 7826.0,
      "output_tokens_median": 72.0,
      "output_tokens_mean": 72.4,
      "output_tokens_min": 72.0,
      "output_tokens_max": 73.0,
      "output_tokens_stdev": 0.5477225575051661,
      "output_tokens_p95": 73.0,
      "prefill_tokens_per_second_median": 2880.263843210583,
      "prefill_tokens_per_second_mean": 2886.8105365675156,
      "prefill_tokens_per_second_min": 2874.6408994621343,
      "prefill_tokens_per_second_max": 2907.0199478638756,
      "prefill_tokens_per_second_stdev": 13.642038834127419,
      "prefill_tokens_per_second_p95": 2907.0199478638756,
      "decode_tokens_per_second_median": 64.2274319761109,
      "decode_tokens_per_second_mean": 64.20668629941896,
      "decode_tokens_per_second_min": 63.63482097403699,
      "decode_tokens_per_second_max": 64.99602802050985,
      "decode_tokens_per_second_stdev": 0.511047776929915,
      "decode_tokens_per_second_p95": 64.99602802050985,
      "load_seconds_median": 0.5716318,
      "load_seconds_mean": 0.57495126,
      "load_seconds_min": 0.5567888,
      "load_seconds_max": 0.5963928,
      "load_seconds_stdev": 0.016294201705760236,
      "load_seconds_p95": 0.5963928,
      "peak_gpu_memory_used_mib_median": 11445.0,
      "peak_system_memory_used_bytes_median": 50285895680.0,
      "gpu_energy_j_observed_median": 1760.9491747415066
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 16384,
      "n": 5,
      "wall_seconds_median": 8.56556359998649,
      "wall_seconds_mean": 8.666807460004929,
      "wall_seconds_min": 8.388355900009628,
      "wall_seconds_max": 8.98048149998067,
      "wall_seconds_stdev": 0.25218584436175207,
      "wall_seconds_p95": 8.98048149998067,
      "ttft_seconds_median": 7.3863080000155605,
      "ttft_seconds_mean": 7.493321900011506,
      "ttft_seconds_min": 7.197479900030885,
      "ttft_seconds_max": 7.80482119997032,
      "ttft_seconds_stdev": 0.26266868460334636,
      "ttft_seconds_p95": 7.80482119997032,
      "prompt_tokens_median": 15966.0,
      "prompt_tokens_mean": 15965.4,
      "prompt_tokens_min": 15964.0,
      "prompt_tokens_max": 15966.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 15966.0,
      "output_tokens_median": 71.0,
      "output_tokens_mean": 71.0,
      "output_tokens_min": 70.0,
      "output_tokens_max": 72.0,
      "output_tokens_stdev": 1.0,
      "output_tokens_p95": 72.0,
      "prefill_tokens_per_second_median": 2931.371105412957,
      "prefill_tokens_per_second_mean": 2912.7548122520793,
      "prefill_tokens_per_second_min": 2853.3362979623776,
      "prefill_tokens_per_second_max": 2951.4972770981744,
      "prefill_tokens_per_second_stdev": 42.67452287761288,
      "prefill_tokens_per_second_p95": 2951.4972770981744,
      "decode_tokens_per_second_median": 61.03685928736851,
      "decode_tokens_per_second_mean": 60.56653289463804,
      "decode_tokens_per_second_min": 59.42774477651347,
      "decode_tokens_per_second_max": 61.398975660422735,
      "decode_tokens_per_second_stdev": 0.9391535961743639,
      "decode_tokens_per_second_p95": 61.398975660422735,
      "load_seconds_median": 0.5853953,
      "load_seconds_mean": 0.57641438,
      "load_seconds_min": 0.5116205,
      "load_seconds_max": 0.614046,
      "load_seconds_stdev": 0.03837861171677525,
      "load_seconds_p95": 0.614046,
      "peak_gpu_memory_used_mib_median": 11978.0,
      "peak_system_memory_used_bytes_median": 51964956672.0,
      "gpu_energy_j_observed_median": 3085.3840927076344
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 32768,
      "n": 5,
      "wall_seconds_median": 16.494066099985503,
      "wall_seconds_mean": 16.408765640005004,
      "wall_seconds_min": 15.760115500015672,
      "wall_seconds_max": 16.71107389999088,
      "wall_seconds_stdev": 0.3750550822791153,
      "wall_seconds_p95": 16.71107389999088,
      "ttft_seconds_median": 15.227572900010273,
      "ttft_seconds_mean": 15.072798440000042,
      "ttft_seconds_min": 14.452683900017291,
      "ttft_seconds_max": 15.28040129999863,
      "ttft_seconds_stdev": 0.3487622329943361,
      "ttft_seconds_p95": 15.28040129999863,
      "prompt_tokens_median": 32244.0,
      "prompt_tokens_mean": 32244.6,
      "prompt_tokens_min": 32244.0,
      "prompt_tokens_max": 32246.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 32246.0,
      "output_tokens_median": 72.0,
      "output_tokens_mean": 73.2,
      "output_tokens_min": 69.0,
      "output_tokens_max": 79.0,
      "output_tokens_stdev": 4.207136793592526,
      "output_tokens_p95": 79.0,
      "prefill_tokens_per_second_median": 2678.9470850623024,
      "prefill_tokens_per_second_mean": 2682.491077486075,
      "prefill_tokens_per_second_min": 2661.8530307760866,
      "prefill_tokens_per_second_max": 2717.16965398598,
      "prefill_tokens_per_second_stdev": 22.215475962320394,
      "prefill_tokens_per_second_p95": 2717.16965398598,
      "decode_tokens_per_second_median": 55.16052074434768,
      "decode_tokens_per_second_mean": 54.87679749180499,
      "decode_tokens_per_second_min": 53.595163572439226,
      "decode_tokens_per_second_max": 55.72515384988127,
      "decode_tokens_per_second_stdev": 0.8227144003339798,
      "decode_tokens_per_second_p95": 55.72515384988127,
      "load_seconds_median": 0.6009852,
      "load_seconds_mean": 0.59675208,
      "load_seconds_min": 0.5605209,
      "load_seconds_max": 0.6357547,
      "load_seconds_stdev": 0.02883680813157033,
      "load_seconds_p95": 0.6357547,
      "peak_gpu_memory_used_mib_median": 12703.0,
      "peak_system_memory_used_bytes_median": 55253917696.0,
      "gpu_energy_j_observed_median": 5857.519420833588
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 15.439900700002909,
      "wall_seconds_mean": 15.26069694000762,
      "wall_seconds_min": 13.74584660003893,
      "wall_seconds_max": 15.961318399989977,
      "wall_seconds_stdev": 0.8826295546479545,
      "wall_seconds_p95": 15.961318399989977,
      "ttft_seconds_median": 7.663781000010204,
      "ttft_seconds_mean": 7.737743760005105,
      "ttft_seconds_min": 7.323985800030641,
      "ttft_seconds_max": 8.363466500013601,
      "ttft_seconds_stdev": 0.42932905910577024,
      "ttft_seconds_p95": 8.363466500013601,
      "prompt_tokens_median": 16342.0,
      "prompt_tokens_mean": 16342.2,
      "prompt_tokens_min": 16341.0,
      "prompt_tokens_max": 16344.0,
      "prompt_tokens_stdev": 1.3038404810405297,
      "prompt_tokens_p95": 16344.0,
      "output_tokens_median": 461.0,
      "output_tokens_mean": 449.0,
      "output_tokens_min": 385.0,
      "output_tokens_max": 504.0,
      "output_tokens_stdev": 47.460509900337144,
      "output_tokens_p95": 504.0,
      "prefill_tokens_per_second_median": 2882.5326494182723,
      "prefill_tokens_per_second_mean": 2881.8162469086583,
      "prefill_tokens_per_second_min": 2782.0161651071444,
      "prefill_tokens_per_second_max": 2963.596781085212,
      "prefill_tokens_per_second_stdev": 77.87477916690666,
      "prefill_tokens_per_second_p95": 2963.596781085212,
      "decode_tokens_per_second_median": 59.76962384745076,
      "decode_tokens_per_second_mean": 59.735967509496156,
      "decode_tokens_per_second_min": 59.29428990843341,
      "decode_tokens_per_second_max": 60.09048625722233,
      "decode_tokens_per_second_stdev": 0.3178824188210302,
      "decode_tokens_per_second_p95": 60.09048625722233,
      "load_seconds_median": 0.5513586,
      "load_seconds_mean": 0.57255808,
      "load_seconds_min": 0.5443296,
      "load_seconds_max": 0.6308371,
      "load_seconds_stdev": 0.037172806026408635,
      "load_seconds_p95": 0.6308371,
      "peak_gpu_memory_used_mib_median": 12229.0,
      "peak_system_memory_used_bytes_median": 52157394944.0,
      "gpu_energy_j_observed_median": 5696.69774291873
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 5.1247525999788195,
      "wall_seconds_mean": 4.965647040004842,
      "wall_seconds_min": 4.39668000000529,
      "wall_seconds_max": 5.342855000053532,
      "wall_seconds_stdev": 0.41331820316183093,
      "wall_seconds_p95": 5.342855000053532,
      "ttft_seconds_median": 3.0642570999916643,
      "ttft_seconds_mean": 3.093392219999805,
      "ttft_seconds_min": 3.0504867000272498,
      "ttft_seconds_max": 3.2230416999664158,
      "ttft_seconds_stdev": 0.072764259231271,
      "ttft_seconds_p95": 3.2230416999664158,
      "prompt_tokens_median": 4148.0,
      "prompt_tokens_mean": 4147.4,
      "prompt_tokens_min": 4146.0,
      "prompt_tokens_max": 4148.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 4148.0,
      "output_tokens_median": 129.0,
      "output_tokens_mean": 118.6,
      "output_tokens_min": 85.0,
      "output_tokens_max": 148.0,
      "output_tokens_stdev": 24.925890154616344,
      "output_tokens_p95": 148.0,
      "prefill_tokens_per_second_median": 2471.905023545074,
      "prefill_tokens_per_second_mean": 2480.6921243763777,
      "prefill_tokens_per_second_min": 2417.429738348781,
      "prefill_tokens_per_second_max": 2546.0614307907704,
      "prefill_tokens_per_second_stdev": 46.69257436410747,
      "prefill_tokens_per_second_p95": 2546.0614307907704,
      "decode_tokens_per_second_median": 62.89951986699835,
      "decode_tokens_per_second_mean": 63.34452354448217,
      "decode_tokens_per_second_min": 62.38794958008459,
      "decode_tokens_per_second_max": 64.89830479488847,
      "decode_tokens_per_second_stdev": 1.0124487887580693,
      "decode_tokens_per_second_p95": 64.89830479488847,
      "load_seconds_median": 0.619134,
      "load_seconds_mean": 0.6127801,
      "load_seconds_min": 0.5545744,
      "load_seconds_max": 0.6889678,
      "load_seconds_stdev": 0.05355983241553318,
      "load_seconds_p95": 0.6889678,
      "peak_gpu_memory_used_mib_median": 11430.0,
      "peak_system_memory_used_bytes_median": 50638450688.0,
      "gpu_energy_j_observed_median": 1682.8222725749017
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 33.88235909998184,
      "wall_seconds_mean": 33.772788040013985,
      "wall_seconds_min": 33.34486390004167,
      "wall_seconds_max": 34.10327680001501,
      "wall_seconds_stdev": 0.3383141413140818,
      "wall_seconds_p95": 34.10327680001501,
      "ttft_seconds_median": 15.634014899958856,
      "ttft_seconds_mean": 15.487217580003199,
      "ttft_seconds_min": 14.8934177000192,
      "ttft_seconds_max": 15.993335000006482,
      "ttft_seconds_stdev": 0.45066414874862515,
      "ttft_seconds_p95": 15.993335000006482,
      "prompt_tokens_median": 32600.0,
      "prompt_tokens_mean": 32599.6,
      "prompt_tokens_min": 32598.0,
      "prompt_tokens_max": 32601.0,
      "prompt_tokens_stdev": 1.140175425099138,
      "prompt_tokens_p95": 32601.0,
      "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": 2650.888620963016,
      "prefill_tokens_per_second_mean": 2661.6044482694692,
      "prefill_tokens_per_second_min": 2645.1767760003154,
      "prefill_tokens_per_second_max": 2683.0900308390746,
      "prefill_tokens_per_second_stdev": 17.814603584403496,
      "prefill_tokens_per_second_p95": 2683.0900308390746,
      "decode_tokens_per_second_median": 54.69606463455836,
      "decode_tokens_per_second_mean": 54.697325113086244,
      "decode_tokens_per_second_min": 54.205392580116516,
      "decode_tokens_per_second_max": 55.224298735623115,
      "decode_tokens_per_second_stdev": 0.3709884347424741,
      "decode_tokens_per_second_p95": 55.224298735623115,
      "load_seconds_median": 0.6318364,
      "load_seconds_mean": 0.6392795,
      "load_seconds_min": 0.547512,
      "load_seconds_max": 0.7533612,
      "load_seconds_stdev": 0.07553356462983855,
      "load_seconds_p95": 0.7533612,
      "peak_gpu_memory_used_mib_median": 12991.0,
      "peak_system_memory_used_bytes_median": 51909218304.0,
      "gpu_energy_j_observed_median": 12683.678114438057
    },
    {
      "system_id": "3090",
      "model": "gemma3:12b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 33.501322600001004,
      "wall_seconds_mean": 33.844878959993366,
      "wall_seconds_min": 32.95163719996344,
      "wall_seconds_max": 35.074595500016585,
      "wall_seconds_stdev": 0.9014366430672999,
      "wall_seconds_p95": 35.074595500016585,
      "ttft_seconds_median": 15.541699900000822,
      "ttft_seconds_mean": 15.536391039995943,
      "ttft_seconds_min": 14.93693099997472,
      "ttft_seconds_max": 16.216645599983167,
      "ttft_seconds_stdev": 0.550862463469368,
      "ttft_seconds_p95": 16.216645599983167,
      "prompt_tokens_median": 32591.0,
      "prompt_tokens_mean": 32590.8,
      "prompt_tokens_min": 32589.0,
      "prompt_tokens_max": 32592.0,
      "prompt_tokens_stdev": 1.3038404810405297,
      "prompt_tokens_p95": 32592.0,
      "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": 2660.929839529406,
      "prefill_tokens_per_second_mean": 2654.9918988298346,
      "prefill_tokens_per_second_min": 2606.6611510068637,
      "prefill_tokens_per_second_max": 2704.0602637634847,
      "prefill_tokens_per_second_stdev": 37.53664141403212,
      "prefill_tokens_per_second_p95": 2704.0602637634847,
      "decode_tokens_per_second_median": 54.73949610651649,
      "decode_tokens_per_second_mean": 54.65476476406066,
      "decode_tokens_per_second_min": 53.075761615948885,
      "decode_tokens_per_second_max": 55.875473684328156,
      "decode_tokens_per_second_stdev": 1.1858216726481101,
      "decode_tokens_per_second_p95": 55.875473684328156,
      "load_seconds_median": 0.6411806,
      "load_seconds_mean": 0.64651982,
      "load_seconds_min": 0.6060859,
      "load_seconds_max": 0.6918602,
      "load_seconds_stdev": 0.03173157707374472,
      "load_seconds_p95": 0.6918602,
      "peak_gpu_memory_used_mib_median": 12968.0,
      "peak_system_memory_used_bytes_median": 52509642752.0,
      "gpu_energy_j_observed_median": 12795.212949161529
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "agent_loop",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 27.931950800004415,
      "wall_seconds_mean": 27.43398042001063,
      "wall_seconds_min": 26.0722452000482,
      "wall_seconds_max": 28.63045619998593,
      "wall_seconds_stdev": 1.2669282898666048,
      "wall_seconds_p95": 28.63045619998593,
      "prompt_tokens_median": 15323.0,
      "prompt_tokens_mean": 15322.8,
      "prompt_tokens_min": 15296.0,
      "prompt_tokens_max": 15358.0,
      "prompt_tokens_stdev": 22.664950915455343,
      "prompt_tokens_p95": 15358.0,
      "output_tokens_median": 544.0,
      "output_tokens_mean": 532.8,
      "output_tokens_min": 495.0,
      "output_tokens_max": 562.0,
      "output_tokens_stdev": 31.76003778335284,
      "output_tokens_p95": 562.0,
      "prefill_tokens_per_second_median": 7496.48946715368,
      "prefill_tokens_per_second_mean": 7497.435627471168,
      "prefill_tokens_per_second_min": 7446.674010458677,
      "prefill_tokens_per_second_max": 7537.639437899928,
      "prefill_tokens_per_second_stdev": 36.42302915491628,
      "prefill_tokens_per_second_p95": 7537.639437899928,
      "decode_tokens_per_second_median": 23.943560681844446,
      "decode_tokens_per_second_mean": 23.939409748052242,
      "decode_tokens_per_second_min": 23.90244137530423,
      "decode_tokens_per_second_max": 23.97208282249971,
      "decode_tokens_per_second_stdev": 0.030149685997465803,
      "decode_tokens_per_second_p95": 23.97208282249971,
      "prefill_seconds_median": 2.040642478,
      "prefill_seconds_mean": 2.0437730276,
      "prefill_seconds_min": 2.032997217,
      "prefill_seconds_max": 2.057697165,
      "prefill_seconds_stdev": 0.009248562110536823,
      "prefill_seconds_p95": 2.057697165,
      "decode_seconds_median": 22.759181435000002,
      "decode_seconds_mean": 22.256264860399998,
      "decode_seconds_min": 20.649019264,
      "decode_seconds_max": 23.471863999999997,
      "decode_seconds_stdev": 1.327419355497639,
      "decode_seconds_p95": 23.471863999999997,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 58920325120.0,
      "gpu_energy_j_observed_median": 1214.901846038103
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 4096,
      "n": 5,
      "wall_seconds_median": 5.303116599970963,
      "wall_seconds_mean": 5.251062819990329,
      "wall_seconds_min": 4.994485000032,
      "wall_seconds_max": 5.362568399985321,
      "wall_seconds_stdev": 0.14761389009056114,
      "wall_seconds_p95": 5.362568399985321,
      "ttft_seconds_median": 2.1304803999955766,
      "ttft_seconds_mean": 2.1370303599978797,
      "ttft_seconds_min": 2.1085351000074297,
      "ttft_seconds_max": 2.17162089998601,
      "ttft_seconds_stdev": 0.02352971792079961,
      "ttft_seconds_p95": 2.17162089998601,
      "prompt_tokens_median": 3754.0,
      "prompt_tokens_mean": 3754.0,
      "prompt_tokens_min": 3753.0,
      "prompt_tokens_max": 3755.0,
      "prompt_tokens_stdev": 0.7071067811865476,
      "prompt_tokens_p95": 3755.0,
      "output_tokens_median": 76.0,
      "output_tokens_mean": 74.4,
      "output_tokens_min": 69.0,
      "output_tokens_max": 76.0,
      "output_tokens_stdev": 3.0495901363953815,
      "output_tokens_p95": 76.0,
      "prefill_tokens_per_second_median": 1929.5290907443048,
      "prefill_tokens_per_second_mean": 1929.8935659795898,
      "prefill_tokens_per_second_min": 1927.2751342804052,
      "prefill_tokens_per_second_max": 1933.5451460037034,
      "prefill_tokens_per_second_stdev": 2.462845050287424,
      "prefill_tokens_per_second_p95": 1933.5451460037034,
      "decode_tokens_per_second_median": 24.422959782243844,
      "decode_tokens_per_second_mean": 24.425276714878756,
      "decode_tokens_per_second_min": 24.39598638567342,
      "decode_tokens_per_second_max": 24.461791494557268,
      "decode_tokens_per_second_stdev": 0.023862126004536625,
      "decode_tokens_per_second_p95": 24.461791494557268,
      "load_seconds_median": 0.1588862,
      "load_seconds_mean": 0.1601791176,
      "load_seconds_min": 0.134295715,
      "load_seconds_max": 0.176928944,
      "load_seconds_stdev": 0.017660964060054546,
      "load_seconds_p95": 0.176928944,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 21752684544.0,
      "gpu_energy_j_observed_median": 314.06043271541597
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 7.624518800002988,
      "wall_seconds_mean": 7.583947059989441,
      "wall_seconds_min": 7.442317099950742,
      "wall_seconds_max": 7.764811199973337,
      "wall_seconds_stdev": 0.1386236701843631,
      "wall_seconds_p95": 7.764811199973337,
      "ttft_seconds_median": 4.3120061000227,
      "ttft_seconds_mean": 4.309781539998949,
      "ttft_seconds_min": 4.277088199974969,
      "ttft_seconds_max": 4.332892600039486,
      "ttft_seconds_stdev": 0.021026826171276325,
      "ttft_seconds_p95": 4.332892600039486,
      "prompt_tokens_median": 7824.0,
      "prompt_tokens_mean": 7823.6,
      "prompt_tokens_min": 7822.0,
      "prompt_tokens_max": 7825.0,
      "prompt_tokens_stdev": 1.140175425099138,
      "prompt_tokens_p95": 7825.0,
      "output_tokens_median": 77.0,
      "output_tokens_mean": 76.0,
      "output_tokens_min": 72.0,
      "output_tokens_max": 81.0,
      "output_tokens_stdev": 3.605551275463989,
      "output_tokens_p95": 81.0,
      "prefill_tokens_per_second_median": 1914.3600673973658,
      "prefill_tokens_per_second_mean": 1913.9336362294475,
      "prefill_tokens_per_second_min": 1909.7892473610646,
      "prefill_tokens_per_second_max": 1917.3945130951588,
      "prefill_tokens_per_second_stdev": 3.1400432924782247,
      "prefill_tokens_per_second_p95": 1917.3945130951588,
      "decode_tokens_per_second_median": 23.587374766218904,
      "decode_tokens_per_second_mean": 23.587098989557955,
      "decode_tokens_per_second_min": 23.57466691108494,
      "decode_tokens_per_second_max": 23.60397592241817,
      "decode_tokens_per_second_stdev": 0.011589523853598422,
      "decode_tokens_per_second_p95": 23.60397592241817,
      "load_seconds_median": 0.171432714,
      "load_seconds_mean": 0.1675103644,
      "load_seconds_min": 0.151681892,
      "load_seconds_max": 0.182520084,
      "load_seconds_stdev": 0.013103130410946108,
      "load_seconds_p95": 0.182520084,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 22281527296.0,
      "gpu_energy_j_observed_median": 487.7271253848076
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 16384,
      "n": 5,
      "wall_seconds_median": 12.217662499984726,
      "wall_seconds_mean": 12.164968939987011,
      "wall_seconds_min": 11.974300400004722,
      "wall_seconds_max": 12.250874199962709,
      "wall_seconds_stdev": 0.11193952107799794,
      "wall_seconds_p95": 12.250874199962709,
      "ttft_seconds_median": 8.731814400001895,
      "ttft_seconds_mean": 8.72760451999493,
      "ttft_seconds_min": 8.700153800018597,
      "ttft_seconds_max": 8.743327100004535,
      "ttft_seconds_stdev": 0.016755571097308488,
      "ttft_seconds_p95": 8.743327100004535,
      "prompt_tokens_median": 15965.0,
      "prompt_tokens_mean": 15964.4,
      "prompt_tokens_min": 15963.0,
      "prompt_tokens_max": 15965.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 15965.0,
      "output_tokens_median": 77.0,
      "output_tokens_mean": 75.6,
      "output_tokens_min": 72.0,
      "output_tokens_max": 77.0,
      "output_tokens_stdev": 2.1908902300206643,
      "output_tokens_p95": 77.0,
      "prefill_tokens_per_second_median": 1889.5147921498765,
      "prefill_tokens_per_second_mean": 1890.246690106256,
      "prefill_tokens_per_second_min": 1887.8982079153755,
      "prefill_tokens_per_second_max": 1896.1107715651588,
      "prefill_tokens_per_second_stdev": 3.389300875916402,
      "prefill_tokens_per_second_p95": 1896.1107715651588,
      "decode_tokens_per_second_median": 22.361357243854002,
      "decode_tokens_per_second_mean": 22.350704365208323,
      "decode_tokens_per_second_min": 22.29923601999757,
      "decode_tokens_per_second_max": 22.397440482400054,
      "decode_tokens_per_second_stdev": 0.038623063508181475,
      "decode_tokens_per_second_p95": 22.397440482400054,
      "load_seconds_median": 0.170054952,
      "load_seconds_mean": 0.1587434626,
      "load_seconds_min": 0.136372752,
      "load_seconds_max": 0.175931581,
      "load_seconds_stdev": 0.01944548577548398,
      "load_seconds_p95": 0.175931581,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 23404138496.0,
      "gpu_energy_j_observed_median": 887.3743614304065
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 32768,
      "n": 5,
      "wall_seconds_median": 22.140859699982684,
      "wall_seconds_mean": 22.11485739998752,
      "wall_seconds_min": 21.89104319998296,
      "wall_seconds_max": 22.233898099977523,
      "wall_seconds_stdev": 0.13194427149392857,
      "wall_seconds_p95": 22.233898099977523,
      "ttft_seconds_median": 18.33773619995918,
      "ttft_seconds_mean": 18.31452079999726,
      "ttft_seconds_min": 18.2486315999995,
      "ttft_seconds_max": 18.345240800001193,
      "ttft_seconds_stdev": 0.040754333630193866,
      "ttft_seconds_p95": 18.345240800001193,
      "prompt_tokens_median": 32243.0,
      "prompt_tokens_mean": 32243.6,
      "prompt_tokens_min": 32243.0,
      "prompt_tokens_max": 32245.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 32245.0,
      "output_tokens_median": 75.0,
      "output_tokens_mean": 75.2,
      "output_tokens_min": 72.0,
      "output_tokens_max": 77.0,
      "output_tokens_stdev": 2.04939015319192,
      "output_tokens_p95": 77.0,
      "prefill_tokens_per_second_median": 1800.3659368088581,
      "prefill_tokens_per_second_mean": 1801.0232137338187,
      "prefill_tokens_per_second_min": 1799.7078833318237,
      "prefill_tokens_per_second_max": 1803.7149335257898,
      "prefill_tokens_per_second_stdev": 1.608183610599127,
      "prefill_tokens_per_second_p95": 1803.7149335257898,
      "decode_tokens_per_second_median": 20.059087760616496,
      "decode_tokens_per_second_mean": 20.060044170105115,
      "decode_tokens_per_second_min": 20.04057233642875,
      "decode_tokens_per_second_max": 20.08046523022695,
      "decode_tokens_per_second_stdev": 0.015162832155933538,
      "decode_tokens_per_second_p95": 20.08046523022695,
      "load_seconds_median": 0.186293524,
      "load_seconds_mean": 0.1789194954,
      "load_seconds_min": 0.134193716,
      "load_seconds_max": 0.204142142,
      "load_seconds_stdev": 0.02631814777897523,
      "load_seconds_p95": 0.204142142,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 25179545600.0,
      "gpu_energy_j_observed_median": 1763.715427210331
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 65536,
      "n": 5,
      "wall_seconds_median": 44.450516499986406,
      "wall_seconds_mean": 44.355595719988926,
      "wall_seconds_min": 43.99572829995304,
      "wall_seconds_max": 44.525265000003856,
      "wall_seconds_stdev": 0.21105886303693283,
      "wall_seconds_p95": 44.525265000003856,
      "ttft_seconds_median": 39.98545689997263,
      "ttft_seconds_mean": 39.94977911999449,
      "ttft_seconds_min": 39.75334070000099,
      "ttft_seconds_max": 40.04904329997953,
      "ttft_seconds_stdev": 0.12020365024241317,
      "ttft_seconds_p95": 40.04904329997953,
      "prompt_tokens_median": 64749.0,
      "prompt_tokens_mean": 64748.6,
      "prompt_tokens_min": 64747.0,
      "prompt_tokens_max": 64750.0,
      "prompt_tokens_stdev": 1.51657508881031,
      "prompt_tokens_p95": 64750.0,
      "output_tokens_median": 74.0,
      "output_tokens_mean": 73.6,
      "output_tokens_min": 70.0,
      "output_tokens_max": 76.0,
      "output_tokens_stdev": 2.6076809620810595,
      "output_tokens_p95": 76.0,
      "prefill_tokens_per_second_median": 1647.5613522262074,
      "prefill_tokens_per_second_mean": 1648.8288777938985,
      "prefill_tokens_per_second_min": 1646.6460150093308,
      "prefill_tokens_per_second_max": 1653.9433806954885,
      "prefill_tokens_per_second_stdev": 2.9886397708249928,
      "prefill_tokens_per_second_p95": 1653.9433806954885,
      "decode_tokens_per_second_median": 16.943152205586994,
      "decode_tokens_per_second_mean": 16.939806364274805,
      "decode_tokens_per_second_min": 16.919108858818714,
      "decode_tokens_per_second_max": 16.95337102427748,
      "decode_tokens_per_second_stdev": 0.014634610000066163,
      "decode_tokens_per_second_p95": 16.95337102427748,
      "load_seconds_median": 0.155838701,
      "load_seconds_mean": 0.1630634198,
      "load_seconds_min": 0.145294082,
      "load_seconds_max": 0.19072207,
      "load_seconds_stdev": 0.019809172096383613,
      "load_seconds_p95": 0.19072207,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 28284604416.0,
      "gpu_energy_j_observed_median": 3737.974726675749
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "context_scaling",
      "context_setting": 131072,
      "n": 5,
      "wall_seconds_median": 98.65267039998434,
      "wall_seconds_mean": 98.71802075997694,
      "wall_seconds_min": 98.58487449999666,
      "wall_seconds_max": 99.00234129995806,
      "wall_seconds_stdev": 0.16419479587461092,
      "wall_seconds_p95": 99.00234129995806,
      "ttft_seconds_median": 92.87623599998187,
      "ttft_seconds_mean": 92.86890469999052,
      "ttft_seconds_min": 92.79717640002491,
      "ttft_seconds_max": 92.91238609998254,
      "ttft_seconds_stdev": 0.04385798439701495,
      "ttft_seconds_p95": 92.91238609998254,
      "prompt_tokens_median": 129815.0,
      "prompt_tokens_mean": 129814.2,
      "prompt_tokens_min": 129813.0,
      "prompt_tokens_max": 129815.0,
      "prompt_tokens_stdev": 1.0954451150103321,
      "prompt_tokens_p95": 129815.0,
      "output_tokens_median": 74.0,
      "output_tokens_mean": 75.0,
      "output_tokens_min": 74.0,
      "output_tokens_max": 79.0,
      "output_tokens_stdev": 2.23606797749979,
      "output_tokens_p95": 79.0,
      "prefill_tokens_per_second_median": 1418.909719657166,
      "prefill_tokens_per_second_mean": 1418.96871175671,
      "prefill_tokens_per_second_min": 1418.4451545897905,
      "prefill_tokens_per_second_max": 1419.5474451564335,
      "prefill_tokens_per_second_stdev": 0.5036844006855922,
      "prefill_tokens_per_second_p95": 1419.5474451564335,
      "decode_tokens_per_second_median": 12.986455620098058,
      "decode_tokens_per_second_mean": 12.98647963637432,
      "decode_tokens_per_second_min": 12.975118970075679,
      "decode_tokens_per_second_max": 12.999150930838635,
      "decode_tokens_per_second_stdev": 0.008722877058857603,
      "decode_tokens_per_second_p95": 12.999150930838635,
      "load_seconds_median": 0.188572795,
      "load_seconds_mean": 0.1864948224,
      "load_seconds_min": 0.172842136,
      "load_seconds_max": 0.198248102,
      "load_seconds_stdev": 0.011904352962838437,
      "load_seconds_p95": 0.198248102,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 35325878272.0,
      "gpu_energy_j_observed_median": 8288.747153700586
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "engineering_validity",
      "context_setting": 20480,
      "n": 5,
      "wall_seconds_median": 30.595761200005654,
      "wall_seconds_mean": 30.439709800004493,
      "wall_seconds_min": 28.157903499959502,
      "wall_seconds_max": 32.322964899998624,
      "wall_seconds_stdev": 1.4893297846329752,
      "wall_seconds_p95": 32.322964899998624,
      "ttft_seconds_median": 9.045793800032698,
      "ttft_seconds_mean": 9.057788720005192,
      "ttft_seconds_min": 9.036147099977825,
      "ttft_seconds_max": 9.10183490003692,
      "ttft_seconds_stdev": 0.026807682780800396,
      "ttft_seconds_p95": 9.10183490003692,
      "prompt_tokens_median": 16341.0,
      "prompt_tokens_mean": 16341.2,
      "prompt_tokens_min": 16340.0,
      "prompt_tokens_max": 16343.0,
      "prompt_tokens_stdev": 1.3038404810405297,
      "prompt_tokens_p95": 16343.0,
      "output_tokens_median": 467.0,
      "output_tokens_mean": 463.6,
      "output_tokens_min": 414.0,
      "output_tokens_max": 505.0,
      "output_tokens_stdev": 32.585272747055534,
      "output_tokens_p95": 505.0,
      "prefill_tokens_per_second_median": 1865.9593806939831,
      "prefill_tokens_per_second_mean": 1864.9931913747523,
      "prefill_tokens_per_second_min": 1859.009009152938,
      "prefill_tokens_per_second_max": 1869.2766510346985,
      "prefill_tokens_per_second_stdev": 3.8794696096942474,
      "prefill_tokens_per_second_p95": 1869.2766510346985,
      "decode_tokens_per_second_median": 21.908803590481785,
      "decode_tokens_per_second_mean": 21.90269750121584,
      "decode_tokens_per_second_min": 21.884336629565517,
      "decode_tokens_per_second_max": 21.920556651674577,
      "decode_tokens_per_second_stdev": 0.01676520097345648,
      "decode_tokens_per_second_p95": 21.920556651674577,
      "load_seconds_median": 0.161482251,
      "load_seconds_mean": 0.173763677,
      "load_seconds_min": 0.155857655,
      "load_seconds_max": 0.206488412,
      "load_seconds_stdev": 0.021267113957342768,
      "load_seconds_p95": 0.206488412,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 60561948672.0,
      "gpu_energy_j_observed_median": 1767.0235483384129
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "interactive",
      "context_setting": 8192,
      "n": 5,
      "wall_seconds_median": 7.435787199996412,
      "wall_seconds_mean": 9.929244880005717,
      "wall_seconds_min": 6.577971300052013,
      "wall_seconds_max": 18.6215741999913,
      "wall_seconds_stdev": 5.101124902583494,
      "wall_seconds_p95": 18.6215741999913,
      "ttft_seconds_median": 2.3737117999698967,
      "ttft_seconds_mean": 3.399745719996281,
      "ttft_seconds_min": 2.334583499992732,
      "ttft_seconds_max": 5.2361119000124745,
      "ttft_seconds_stdev": 1.4364654902286156,
      "ttft_seconds_p95": 5.2361119000124745,
      "prompt_tokens_median": 4147.0,
      "prompt_tokens_mean": 4146.4,
      "prompt_tokens_min": 4145.0,
      "prompt_tokens_max": 4147.0,
      "prompt_tokens_stdev": 0.8944271909999159,
      "prompt_tokens_p95": 4147.0,
      "output_tokens_median": 110.0,
      "output_tokens_mean": 110.4,
      "output_tokens_min": 100.0,
      "output_tokens_max": 121.0,
      "output_tokens_stdev": 10.502380682492898,
      "output_tokens_p95": 121.0,
      "prefill_tokens_per_second_median": 1895.969384242992,
      "prefill_tokens_per_second_mean": 1500.5273687157294,
      "prefill_tokens_per_second_min": 831.4052678620861,
      "prefill_tokens_per_second_max": 1909.1241448091168,
      "prefill_tokens_per_second_stdev": 554.9560413690413,
      "prefill_tokens_per_second_p95": 1909.1241448091168,
      "decode_tokens_per_second_median": 23.991092699721793,
      "decode_tokens_per_second_mean": 20.145323361826872,
      "decode_tokens_per_second_min": 9.122373996319366,
      "decode_tokens_per_second_max": 24.04884297884067,
      "decode_tokens_per_second_stdev": 6.456159159249896,
      "decode_tokens_per_second_p95": 24.04884297884067,
      "load_seconds_median": 0.177169571,
      "load_seconds_mean": 0.1864072498,
      "load_seconds_min": 0.125800947,
      "load_seconds_max": 0.264506391,
      "load_seconds_stdev": 0.05264107515315704,
      "load_seconds_p95": 0.264506391,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 58926747648.0,
      "gpu_energy_j_observed_median": 422.2897865319252
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "large_context",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 70.29821310000261,
      "wall_seconds_mean": 70.29600400000345,
      "wall_seconds_min": 70.27188860002207,
      "wall_seconds_max": 70.32819660002133,
      "wall_seconds_stdev": 0.024352270809909732,
      "wall_seconds_p95": 70.32819660002133,
      "ttft_seconds_median": 18.746604099986143,
      "ttft_seconds_mean": 18.74271836001426,
      "ttft_seconds_min": 18.71749250002904,
      "ttft_seconds_max": 18.76964680000674,
      "ttft_seconds_stdev": 0.021014125576700522,
      "ttft_seconds_p95": 18.76964680000674,
      "prompt_tokens_median": 32599.0,
      "prompt_tokens_mean": 32598.6,
      "prompt_tokens_min": 32597.0,
      "prompt_tokens_max": 32600.0,
      "prompt_tokens_stdev": 1.140175425099138,
      "prompt_tokens_p95": 32600.0,
      "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": 1776.961495496446,
      "prefill_tokens_per_second_mean": 1776.5871438486968,
      "prefill_tokens_per_second_min": 1774.7959665547698,
      "prefill_tokens_per_second_max": 1777.373564128234,
      "prefill_tokens_per_second_stdev": 1.0242298158285021,
      "prefill_tokens_per_second_p95": 1777.373564128234,
      "decode_tokens_per_second_median": 19.563171658195802,
      "decode_tokens_per_second_mean": 19.561175904227817,
      "decode_tokens_per_second_min": 19.551370066724914,
      "decode_tokens_per_second_max": 19.568187353024932,
      "decode_tokens_per_second_stdev": 0.006263754087129324,
      "decode_tokens_per_second_p95": 19.568187353024932,
      "load_seconds_median": 0.159060134,
      "load_seconds_mean": 0.1620075106,
      "load_seconds_min": 0.137114803,
      "load_seconds_max": 0.184287242,
      "load_seconds_stdev": 0.01771652148970104,
      "load_seconds_p95": 0.184287242,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 62327312384.0,
      "gpu_energy_j_observed_median": 4044.7319762408747
    },
    {
      "system_id": "spark",
      "model": "gemma3:12b",
      "workload": "overnight",
      "context_setting": 36864,
      "n": 5,
      "wall_seconds_median": 70.37869680003496,
      "wall_seconds_mean": 70.34962903999258,
      "wall_seconds_min": 70.0851224999642,
      "wall_seconds_max": 70.56000569998287,
      "wall_seconds_stdev": 0.17962229429041415,
      "wall_seconds_p95": 70.56000569998287,
      "ttft_seconds_median": 18.74087579996558,
      "ttft_seconds_mean": 18.756515879987273,
      "ttft_seconds_min": 18.691443699994124,
      "ttft_seconds_max": 18.817869599966798,
      "ttft_seconds_stdev": 0.05707155342450909,
      "ttft_seconds_p95": 18.817869599966798,
      "prompt_tokens_median": 32590.0,
      "prompt_tokens_mean": 32589.8,
      "prompt_tokens_min": 32588.0,
      "prompt_tokens_max": 32591.0,
      "prompt_tokens_stdev": 1.3038404810405297,
      "prompt_tokens_p95": 32591.0,
      "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": 1776.7527025230386,
      "prefill_tokens_per_second_mean": 1777.319263633988,
      "prefill_tokens_per_second_min": 1773.050924230411,
      "prefill_tokens_per_second_max": 1783.9279349162146,
      "prefill_tokens_per_second_stdev": 4.520464013839378,
      "prefill_tokens_per_second_p95": 1783.9279349162146,
      "decode_tokens_per_second_median": 19.55115304019137,
      "decode_tokens_per_second_mean": 19.546648337017043,
      "decode_tokens_per_second_min": 19.461277243363533,
      "decode_tokens_per_second_max": 19.614623419128506,
      "decode_tokens_per_second_stdev": 0.05482410881081344,
      "decode_tokens_per_second_p95": 19.614623419128506,
      "load_seconds_median": 0.155217221,
      "load_seconds_mean": 0.163550825,
      "load_seconds_min": 0.130536607,
      "load_seconds_max": 0.189441795,
      "load_seconds_stdev": 0.02487498037220453,
      "load_seconds_p95": 0.189441795,
      "peak_gpu_memory_used_mib_median": null,
      "peak_system_memory_used_bytes_median": 62186553344.0,
      "gpu_energy_j_observed_median": 3995.6306132781506
    }
  ]
}
