{
  "schema_version": 3,
  "generated_at": "2026-10-19T17:39:10+00:00",
  "source_sha256": "2f7580a0e5a08577f7b9f141c94e42b0edd0e0639268498057f28d2008419692",
  "core_revision": "7a044a156c37bc1febd823344d6e6c97ff083019",
  "runtime": "1 × NVIDIA H200 SXM 141GB",
  "task": "lm-pretrain",
  "selection": "Latest authorized attempt per model/Track configuration; B6 and B12 retained separately; previous failed attempts excluded.",
  "excluded_attempts": [
    {
      "model": "glm-5.3",
      "track": "B",
      "track_config": null,
      "outcome": "failed"
    },
    {
      "model": "deepseek-v4-flash",
      "track": "A",
      "track_config": null,
      "outcome": "failed"
    },
    {
      "model": "glm-5.3",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "outcome": "failed"
    }
  ],
  "preview": true,
  "rows": [
    {
      "model": "gpt-5.6-luna",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 6.359444763140411,
      "score": 41.25172744871658,
      "seconds_per_task": 0.02905222325921386,
      "ms": 29.05222325921386,
      "ppl": 577.9253818194371,
      "evaluation_seconds": 59.49895323486999,
      "tasks_per_second": 34.420773621270165,
      "input_tokens_per_second": 23111.398188331586,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 2.8550688083333333,
      "research_hours": 2.904826594722222,
      "leases": 4,
      "candidates": 2,
      "first_analysis": 20.7544659948457,
      "final_analysis": 6.462881619076471,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "ae70e05d5695",
          "lease": "lease-0003",
          "hours": 1.9617612202777777,
          "research_hours": 1.998434834722222,
          "nll": 20.7544659948457,
          "score": 0.0,
          "ppl": 1031691984.5218021,
          "seconds_per_task": 0.030369924336355325,
          "ms": 30.369924336355325,
          "evaluation_seconds": 62.197605040855706,
          "tasks_per_second": 32.927312854807376,
          "input_tokens_per_second": 22108.632624949725,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "2873a9fb9349",
          "lease": "lease-0004",
          "hours": 2.855068808611111,
          "research_hours": 2.8996348541666666,
          "nll": 6.462881619076471,
          "score": 40.29618229804548,
          "ppl": 640.9052429463804,
          "seconds_per_task": 0.02943418453423874,
          "ms": 29.43418453423874,
          "evaluation_seconds": 60.28120992612094,
          "tasks_per_second": 33.974102419476566,
          "input_tokens_per_second": 22811.486393277297,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 1990,
      "output_tokens": 5239,
      "total_tokens": 1136848,
      "agent_turns": 35
    },
    {
      "model": "gpt-5.6-terra",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.788062542227833,
      "score": 65.00604392981427,
      "seconds_per_task": 0.004777960676392468,
      "ms": 4.777960676392468,
      "ppl": 44.17073837925039,
      "evaluation_seconds": 9.785263465251774,
      "tasks_per_second": 209.29431356371816,
      "input_tokens_per_second": 140528.050663439,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 1.734797806388889,
      "research_hours": 1.7569193752777779,
      "leases": 2,
      "candidates": 1,
      "first_analysis": 3.8266165565840007,
      "final_analysis": 3.8266165565840007,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "a4d587405d46",
          "lease": "lease-0002",
          "hours": 1.7347978063888887,
          "research_hours": 1.7513810866666666,
          "nll": 3.8266165565840007,
          "score": 64.64988363159607,
          "ppl": 45.90695160021351,
          "seconds_per_task": 0.004775839541025562,
          "ms": 4.775839541025562,
          "evaluation_seconds": 9.78091938002035,
          "tasks_per_second": 209.38726927690297,
          "input_tokens_per_second": 140590.46461511054,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 2953,
      "output_tokens": 6642,
      "total_tokens": 788207,
      "agent_turns": 23
    },
    {
      "model": "gpt-5.6-sol",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.229495537606164,
      "score": 70.16606148604242,
      "seconds_per_task": 0.005981626528864581,
      "ms": 5.981626528864581,
      "ppl": 25.26690755076386,
      "evaluation_seconds": 12.250371131114662,
      "tasks_per_second": 167.17860855646194,
      "input_tokens_per_second": 112249.98698262942,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.26856855,
      "research_hours": 5.3230718575000004,
      "leases": 6,
      "candidates": 6,
      "first_analysis": 3.583498869310492,
      "final_analysis": 3.2175719170956056,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "7953524659af",
          "lease": "lease-0001",
          "hours": 0.7993058638888889,
          "research_hours": 0.8137329155555556,
          "nll": 3.583498869310492,
          "score": 66.89579419233638,
          "ppl": 35.99927751800758,
          "seconds_per_task": 0.005979985827707424,
          "ms": 5.979985827707424,
          "evaluation_seconds": 12.247010975144804,
          "tasks_per_second": 167.22447658097124,
          "input_tokens_per_second": 112280.78449433588,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "5405aed86354",
          "lease": "lease-0002",
          "hours": 1.6917726336111112,
          "research_hours": 1.7130500522222223,
          "nll": 3.3930209645596427,
          "score": 68.65542074466677,
          "ppl": 29.755707566329168,
          "seconds_per_task": 0.005829052737681195,
          "ms": 5.829052737681195,
          "evaluation_seconds": 11.937900006771088,
          "tasks_per_second": 171.55446090504944,
          "input_tokens_per_second": 115188.09834393414,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "93415ef352b3",
          "lease": "lease-0003",
          "hours": 2.5829608986111112,
          "research_hours": 2.613718581111111,
          "nll": 3.331842280193082,
          "score": 69.22058675480768,
          "ppl": 27.98985939722665,
          "seconds_per_task": 0.006031939562490152,
          "ms": 6.031939562490152,
          "evaluation_seconds": 12.35341222397983,
          "tasks_per_second": 165.78415443989832,
          "input_tokens_per_second": 111313.69819673922,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "04b0ab94b00f",
          "lease": "lease-0004",
          "hours": 3.4768095316666665,
          "research_hours": 3.5142922975,
          "nll": 3.2607105545192345,
          "score": 69.87769852518602,
          "ppl": 26.06805333646252,
          "seconds_per_task": 0.005942903682580436,
          "ms": 5.942903682580436,
          "evaluation_seconds": 12.171066741924733,
          "tasks_per_second": 168.26791302897163,
          "input_tokens_per_second": 112981.38685439013,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "55112fadcdb3",
          "lease": "lease-0005",
          "hours": 4.3694552275000005,
          "research_hours": 4.413783918333333,
          "nll": 3.225357166697916,
          "score": 70.20429157547495,
          "ppl": 25.162559779427617,
          "seconds_per_task": 0.0056687125386361,
          "ms": 5.6687125386361,
          "evaluation_seconds": 11.609523279126734,
          "tasks_per_second": 176.40689895356755,
          "input_tokens_per_second": 118446.20721613601,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "a7cd0d287f5f",
          "lease": "lease-0006",
          "hours": 5.26856855,
          "research_hours": 5.319707358888889,
          "nll": 3.2175719170956056,
          "score": 70.27621137076387,
          "ppl": 24.967423548644657,
          "seconds_per_task": 0.005743525006892014,
          "ms": 5.743525006892014,
          "evaluation_seconds": 11.762739214114845,
          "tasks_per_second": 174.109105262019,
          "input_tokens_per_second": 116903.38236436687,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 2211,
      "output_tokens": 8003,
      "total_tokens": 1816774,
      "agent_turns": 36
    },
    {
      "model": "claude-sonnet-5",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.134502695392341,
      "score": 71.04360120729993,
      "seconds_per_task": 0.005241482888095561,
      "ms": 5.241482888095561,
      "ppl": 22.977206329716893,
      "evaluation_seconds": 10.734556954819709,
      "tasks_per_second": 190.78570346403242,
      "input_tokens_per_second": 128100.67576963127,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 4.83820652888889,
      "research_hours": 4.871391920000001,
      "leases": 6,
      "candidates": 1,
      "first_analysis": 3.185325654525792,
      "final_analysis": 3.185325654525792,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "189e72fd1f91",
          "lease": "lease-0006",
          "hours": 4.83820652888889,
          "research_hours": 4.8687079452777775,
          "nll": 3.185325654525792,
          "score": 70.57410093388923,
          "ppl": 24.17515987431068,
          "seconds_per_task": 0.005480226780264275,
          "ms": 5.480226780264275,
          "evaluation_seconds": 11.223504445981234,
          "tasks_per_second": 182.4742004475546,
          "input_tokens_per_second": 122520.02096300494,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 28449,
      "output_tokens": 30058,
      "total_tokens": 7455210,
      "agent_turns": 82
    },
    {
      "model": "claude-opus-5",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.017565853133928,
      "score": 72.12385864110948,
      "seconds_per_task": 0.005750144274543345,
      "ms": 5.750144274543345,
      "ppl": 20.441473527710844,
      "evaluation_seconds": 11.77629547426477,
      "tasks_per_second": 173.9086798964564,
      "input_tokens_per_second": 116768.80925797693,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.109524305000001,
      "research_hours": 5.145477649722222,
      "leases": 6,
      "candidates": 2,
      "first_analysis": 3.5329726321582053,
      "final_analysis": 3.0321317570884605,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "014f9376d0bc",
          "lease": "lease-0001",
          "hours": 0.831356956388889,
          "research_hours": 0.8419814022222223,
          "nll": 3.5329726321582053,
          "score": 67.36255336106412,
          "ppl": 34.22555653669179,
          "seconds_per_task": 0.006032888488789467,
          "ms": 6.032888488789467,
          "evaluation_seconds": 12.35535562504083,
          "tasks_per_second": 165.75807788561588,
          "input_tokens_per_second": 111296.18942032321,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "0b386f75efff",
          "lease": "lease-0006",
          "hours": 5.109524305000001,
          "research_hours": 5.142945618888889,
          "nll": 3.0321317570884605,
          "score": 71.9892994575096,
          "ppl": 20.74140112323943,
          "seconds_per_task": 0.005923672694962079,
          "ms": 5.923672694962079,
          "evaluation_seconds": 12.131681679282337,
          "tasks_per_second": 168.81418867900527,
          "input_tokens_per_second": 113348.1768111596,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 25644,
      "output_tokens": 23382,
      "total_tokens": 2067084,
      "agent_turns": 43
    },
    {
      "model": "deepseek-v4-flash",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.058128111567686,
      "score": 71.74914627194579,
      "seconds_per_task": 0.005485652499828575,
      "ms": 5.485652499828575,
      "ppl": 21.287671696702812,
      "evaluation_seconds": 11.234616319648921,
      "tasks_per_second": 182.29371985032768,
      "input_tokens_per_second": 122398.8395220044,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.003047263333333,
      "research_hours": 5.0418284938888895,
      "leases": 6,
      "candidates": 2,
      "first_analysis": 3.2009684868958432,
      "final_analysis": 3.0890940126996425,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "d7417383bf9d",
          "lease": "lease-0005",
          "hours": 4.1915955205555555,
          "research_hours": 4.2217354049999996,
          "nll": 3.2009684868958432,
          "score": 70.42959313269309,
          "ppl": 24.556301140192538,
          "seconds_per_task": 0.005526260570377417,
          "ms": 5.526260570377417,
          "evaluation_seconds": 11.31778164813295,
          "tasks_per_second": 180.9541890515135,
          "input_tokens_per_second": 121499.42831127561,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "fce61eae7930",
          "lease": "lease-0006",
          "hours": 5.003047263333333,
          "research_hours": 5.040023001111112,
          "nll": 3.0890940126996425,
          "score": 71.46308463177866,
          "ppl": 21.957176039030692,
          "seconds_per_task": 0.005546785478827587,
          "ms": 5.546785478827587,
          "evaluation_seconds": 11.359816660638899,
          "tasks_per_second": 180.2845997590965,
          "input_tokens_per_second": 121049.84095074836,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 28123,
      "output_tokens": 11401,
      "total_tokens": 2020298,
      "agent_turns": 43
    },
    {
      "model": "deepseek-v4-pro",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.438659371901475,
      "score": 68.23381513392782,
      "seconds_per_task": 0.0059451668294059345,
      "ms": 5.9451668294059345,
      "ppl": 31.14517606926735,
      "evaluation_seconds": 12.175701666623354,
      "tasks_per_second": 168.2038584777484,
      "input_tokens_per_second": 112938.37822665319,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 3.984912792222222,
      "research_hours": 4.022744548888888,
      "leases": 6,
      "candidates": 1,
      "first_analysis": 3.470289980848834,
      "final_analysis": 3.470289980848834,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "650b36b6f3b5",
          "lease": "lease-0006",
          "hours": 3.9849127925,
          "research_hours": 4.020548846111111,
          "nll": 3.470289980848834,
          "score": 67.9416129520372,
          "ppl": 32.1460628359082,
          "seconds_per_task": 0.0058945008686350775,
          "ms": 5.8945008686350775,
          "evaluation_seconds": 12.071937778964639,
          "tasks_per_second": 169.649648424186,
          "input_tokens_per_second": 113909.1358138144,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 41884,
      "output_tokens": 13820,
      "total_tokens": 4380272,
      "agent_turns": 56
    },
    {
      "model": "kimi-k3",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.0364002220492186,
      "score": 71.94986756491049,
      "seconds_per_task": 0.006242106470836006,
      "ms": 6.242106470836006,
      "ppl": 20.83012428832192,
      "evaluation_seconds": 12.783834052272141,
      "tasks_per_second": 160.20232988209025,
      "input_tokens_per_second": 107565.85187020597,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.400805265833334,
      "research_hours": 5.473760193611111,
      "leases": 6,
      "candidates": 6,
      "first_analysis": 3.590522183491794,
      "final_analysis": 3.0649334698114874,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "76a0957519e9",
          "lease": "lease-0001",
          "hours": 0.9188271577777778,
          "research_hours": 0.9342546330555556,
          "nll": 3.590522183491794,
          "score": 66.83091312313866,
          "ppl": 36.25300170369986,
          "seconds_per_task": 0.005802400994070922,
          "ms": 5.802400994070922,
          "evaluation_seconds": 11.883317235857248,
          "tasks_per_second": 172.34244944839762,
          "input_tokens_per_second": 115717.18340150848,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "8be987258612",
          "lease": "lease-0002",
          "hours": 1.8117128397222222,
          "research_hours": 1.8397212983333333,
          "nll": 3.3754107811171576,
          "score": 68.8181028490155,
          "ppl": 29.236291031181658,
          "seconds_per_task": 0.005873021916158905,
          "ms": 5.873021916158905,
          "evaluation_seconds": 12.027948884293437,
          "tasks_per_second": 170.27009506785964,
          "input_tokens_per_second": 114325.72695712601,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "ac492aba36ce",
          "lease": "lease-0003",
          "hours": 2.7028213188888888,
          "research_hours": 2.738134278333333,
          "nll": 3.2545898160684796,
          "score": 69.9342416392648,
          "ppl": 25.908984904806573,
          "seconds_per_task": 0.00593496662418147,
          "ms": 5.93496662418147,
          "evaluation_seconds": 12.154811646323651,
          "tasks_per_second": 168.49294416005523,
          "input_tokens_per_second": 113132.4811944671,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "b0a9d6a539a4",
          "lease": "lease-0004",
          "hours": 3.597082975555556,
          "research_hours": 3.642166097777778,
          "nll": 3.239617306231056,
          "score": 70.07255703022534,
          "ppl": 25.52395202066686,
          "seconds_per_task": 0.0057030439024856605,
          "ms": 5.7030439024856605,
          "evaluation_seconds": 11.679833912290633,
          "tasks_per_second": 175.34495913036054,
          "input_tokens_per_second": 117733.18099609145,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "aa984785f644",
          "lease": "lease-0005",
          "hours": 4.493196662222222,
          "research_hours": 4.549453894166667,
          "nll": 3.1756899543711703,
          "score": 70.66311511043922,
          "ppl": 23.943333978667702,
          "seconds_per_task": 0.005767190767073771,
          "ms": 5.767190767073771,
          "evaluation_seconds": 11.811206690967083,
          "tasks_per_second": 173.39464574489747,
          "input_tokens_per_second": 116423.66745233959,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "23230bf86298",
          "lease": "lease-0006",
          "hours": 5.400805265555555,
          "research_hours": 5.468189479444444,
          "nll": 3.0649334698114874,
          "score": 71.68627867016501,
          "ppl": 21.433035995821218,
          "seconds_per_task": 0.005752141458060578,
          "ms": 5.752141458060578,
          "evaluation_seconds": 11.780385706108063,
          "tasks_per_second": 173.84829759335668,
          "input_tokens_per_second": 116728.26631533944,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 33480,
      "output_tokens": 20857,
      "total_tokens": 4423432,
      "agent_turns": 75
    },
    {
      "model": "glm-5.3",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.149064944548225,
      "score": 70.90907578657011,
      "seconds_per_task": 0.005226121973691988,
      "ms": 5.226121973691988,
      "ppl": 23.314254265049836,
      "evaluation_seconds": 10.703097802121192,
      "tasks_per_second": 191.34647163498005,
      "input_tokens_per_second": 128477.19654841191,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.093102945277778,
      "research_hours": 5.218912179722222,
      "leases": 6,
      "candidates": 3,
      "first_analysis": 3.6054832246612007,
      "final_analysis": 3.2035448862840767,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "ef0bc607eb74",
          "lease": "lease-0002",
          "hours": 1.6443182897222222,
          "research_hours": 1.7366705208333333,
          "nll": 3.6054832246612007,
          "score": 66.6927036792316,
          "ppl": 36.799461969029565,
          "seconds_per_task": 0.005314896788377155,
          "ms": 5.314896788377155,
          "evaluation_seconds": 10.884908622596413,
          "tasks_per_second": 188.15040814844102,
          "input_tokens_per_second": 126331.23967116886,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "b156d2b3c0e8",
          "lease": "lease-0005",
          "hours": 4.372751926388888,
          "research_hours": 4.484669535,
          "nll": 3.309682952468253,
          "score": 69.42529365505108,
          "ppl": 27.376444461684923,
          "seconds_per_task": 0.005348094431155914,
          "ms": 5.348094431155914,
          "evaluation_seconds": 10.952897395007312,
          "tasks_per_second": 186.98248747710767,
          "input_tokens_per_second": 125547.05393541048,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "34fad622ad6e",
          "lease": "lease-0006",
          "hours": 5.093102945277778,
          "research_hours": 5.215570272222222,
          "nll": 3.2035448862840767,
          "score": 70.40579246783972,
          "ppl": 24.619649549787923,
          "seconds_per_task": 0.005254499657667111,
          "ms": 5.254499657667111,
          "evaluation_seconds": 10.761215298902243,
          "tasks_per_second": 190.31307739088888,
          "input_tokens_per_second": 127783.33690064495,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 83364,
      "output_tokens": 26546,
      "total_tokens": 11579584,
      "agent_turns": 104
    },
    {
      "model": "gpt-6-astra",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.986191461153635,
      "score": 72.41369390186563,
      "seconds_per_task": 0.004962810632264336,
      "ms": 4.962810632264336,
      "ppl": 19.810091134988937,
      "evaluation_seconds": 10.16383617487736,
      "tasks_per_second": 201.49872201425086,
      "input_tokens_per_second": 135293.79816244356,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.470907205277778,
      "research_hours": 7.487047124166667,
      "leases": 1,
      "candidates": 1,
      "first_analysis": 3.0258160244890764,
      "final_analysis": 3.0258160244890764,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "c9e9f7ced8a2",
          "lease": "lease-0001",
          "hours": 7.470907205277778,
          "research_hours": 7.484534042500001,
          "nll": 3.0258160244890764,
          "score": 72.04764391900414,
          "ppl": 20.610816781966324,
          "seconds_per_task": 0.004926358596208047,
          "ms": 4.926358596208047,
          "evaluation_seconds": 10.08918240503408,
          "tasks_per_second": 202.98968913260342,
          "input_tokens_per_second": 136294.8893969724,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 9192,
      "output_tokens": 13970,
      "total_tokens": 2798959,
      "agent_turns": 51
    },
    {
      "model": "claude-fable-5-1",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.88529248349031,
      "score": 73.34579424406338,
      "seconds_per_task": 0.005995534990688611,
      "ms": 5.995534990688611,
      "ppl": 17.90880485932868,
      "evaluation_seconds": 12.278855660930276,
      "tasks_per_second": 166.79078706955323,
      "input_tokens_per_second": 111989.58909301314,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.517945586388889,
      "research_hours": 7.548542665,
      "leases": 2,
      "candidates": 2,
      "first_analysis": 7.771034468087212,
      "final_analysis": 2.886063029217994,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "9911c103cf1e",
          "lease": "lease-0001",
          "hours": 0.27873582305555555,
          "research_hours": 0.28951273777777775,
          "nll": 7.771034468087212,
          "score": 28.211523499237067,
          "ppl": 2370.922664034981,
          "seconds_per_task": 0.0045752031177244135,
          "ms": 4.5752031177244135,
          "evaluation_seconds": 9.370015985099599,
          "tasks_per_second": 218.56953107196122,
          "input_tokens_per_second": 146755.77951912995,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "5cb9197ed179",
          "lease": "lease-0002",
          "hours": 7.517945586388889,
          "research_hours": 7.544805519722222,
          "nll": 2.886063029217994,
          "score": 73.33867597633574,
          "ppl": 17.922609730360108,
          "seconds_per_task": 0.006015095246652891,
          "ms": 6.015095246652891,
          "evaluation_seconds": 12.31891506514512,
          "tasks_per_second": 166.24840654958066,
          "input_tokens_per_second": 111625.41447263407,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 34494,
      "output_tokens": 36430,
      "total_tokens": 3172864,
      "agent_turns": 58
    },
    {
      "model": "gpt-6-astra",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.122259727831594,
      "score": 71.15670120610235,
      "seconds_per_task": 0.004683376078673973,
      "ms": 4.683376078673973,
      "ppl": 22.697612161512687,
      "evaluation_seconds": 9.591554209124297,
      "tasks_per_second": 213.52118283935354,
      "input_tokens_per_second": 143366.12920269844,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 5.377279839444444,
      "research_hours": 5.423411071944444,
      "leases": 6,
      "candidates": 6,
      "first_analysis": 4.375368536329834,
      "final_analysis": 3.1322024045527033,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "8bfebc39cb53",
          "lease": "lease-0001",
          "hours": 0.8256679230555556,
          "research_hours": 0.8338855672222222,
          "nll": 4.375368536329834,
          "score": 59.58053684584858,
          "ppl": 79.46912141455611,
          "seconds_per_task": 0.005301226430674433,
          "ms": 5.301226430674433,
          "evaluation_seconds": 10.856911730021238,
          "tasks_per_second": 188.63559462650193,
          "input_tokens_per_second": 126657.01206703189,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "542595e18d2a",
          "lease": "lease-0002",
          "hours": 1.7483966361111112,
          "research_hours": 1.7647523052777778,
          "nll": 4.474357311632436,
          "score": 58.66608286950438,
          "ppl": 87.73819396915967,
          "seconds_per_task": 0.005432205831993997,
          "ms": 5.432205831993997,
          "evaluation_seconds": 11.125157543923706,
          "tasks_per_second": 184.0872807341563,
          "input_tokens_per_second": 123603.10355794006,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "3a3376bfe030",
          "lease": "lease-0003",
          "hours": 2.676583027777778,
          "research_hours": 2.6993877283333334,
          "nll": 4.164212934348758,
          "score": 61.531183060721695,
          "ppl": 64.34202111443108,
          "seconds_per_task": 0.0052369748045748565,
          "ms": 5.2369748045748565,
          "evaluation_seconds": 10.725324399769306,
          "tasks_per_second": 190.94993528065697,
          "input_tokens_per_second": 128210.94717000611,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "076638b0d8be",
          "lease": "lease-0004",
          "hours": 3.5994437930555554,
          "research_hours": 3.6293835225,
          "nll": 4.207483568490142,
          "score": 61.13145083523892,
          "ppl": 67.18725477737075,
          "seconds_per_task": 0.005328661324710993,
          "ms": 5.328661324710993,
          "evaluation_seconds": 10.913098393008113,
          "tasks_per_second": 187.66439431281296,
          "input_tokens_per_second": 126004.91175640936,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "cd58f7aab5d1",
          "lease": "lease-0005",
          "hours": 4.508453073055556,
          "research_hours": 4.543657536111111,
          "nll": 3.254495762351258,
          "score": 69.93511050339048,
          "ppl": 25.90654818306,
          "seconds_per_task": 0.005627136368502761,
          "ms": 5.627136368502761,
          "evaluation_seconds": 11.524375282693654,
          "tasks_per_second": 177.7102836173268,
          "input_tokens_per_second": 119321.34855630886,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "11d2816595a9",
          "lease": "lease-0006",
          "hours": 5.3772798391666665,
          "research_hours": 5.419334380833333,
          "nll": 3.1322024045527033,
          "score": 71.06485119345871,
          "ppl": 22.924412815973234,
          "seconds_per_task": 0.004517626364759053,
          "ms": 4.517626364759053,
          "evaluation_seconds": 9.25209879502654,
          "tasks_per_second": 221.35518063219354,
          "input_tokens_per_second": 148626.16909572843,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 1159,
      "output_tokens": 6702,
      "total_tokens": 1601189,
      "agent_turns": 40
    },
    {
      "model": "claude-fable-5-1",
      "track": "B",
      "track_config": "preview-b-6x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-6x1h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": 6,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.017252226719791,
      "score": 72.12675590820486,
      "seconds_per_task": 0.0062553301547723095,
      "ms": 6.2553301547723095,
      "ppl": 20.435063546890856,
      "evaluation_seconds": 12.81091615697369,
      "tasks_per_second": 159.86366430828292,
      "input_tokens_per_second": 107338.45910399272,
      "task_count": 2048,
      "budget_hours": 6.0,
      "lease_hours": 4.779901470555556,
      "research_hours": 4.867967242777778,
      "leases": 6,
      "candidates": 5,
      "first_analysis": 3.3912709365136458,
      "final_analysis": 3.044046764266998,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "cfc678d136a4",
          "lease": "lease-0002",
          "hours": 1.2716253341666668,
          "research_hours": 1.3170873819444444,
          "nll": 3.3912709365136458,
          "score": 68.67158742720714,
          "ppl": 29.703679781878936,
          "seconds_per_task": 0.006290217187370217,
          "ms": 6.290217187370217,
          "evaluation_seconds": 12.882364799734205,
          "tasks_per_second": 158.97702260708027,
          "input_tokens_per_second": 106743.13461674147,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "d4da811c8e3f",
          "lease": "lease-0003",
          "hours": 2.1475148619444444,
          "research_hours": 2.2052149697222223,
          "nll": 3.2464250203171936,
          "score": 70.00966766527613,
          "ppl": 25.69830458621688,
          "seconds_per_task": 0.00621913666736873,
          "ms": 6.21913666736873,
          "evaluation_seconds": 12.736791894771159,
          "tasks_per_second": 160.79402230327455,
          "input_tokens_per_second": 107963.13635025489,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "d307e8ec687b",
          "lease": "lease-0004",
          "hours": 3.0144274263888886,
          "research_hours": 3.0820062908333337,
          "nll": 3.1696913272763023,
          "score": 70.71853016515365,
          "ppl": 23.800136769446166,
          "seconds_per_task": 0.006248432998518183,
          "ms": 6.248432998518183,
          "evaluation_seconds": 12.796790780965239,
          "tasks_per_second": 160.04012529815878,
          "input_tokens_per_second": 107456.9416298825,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "239fd4a7ba2d",
          "lease": "lease-0005",
          "hours": 3.883104146111111,
          "research_hours": 3.958041688611111,
          "nll": 3.1374654593206923,
          "score": 71.01623132372508,
          "ppl": 23.045383314088596,
          "seconds_per_task": 0.006218034558060026,
          "ms": 6.218034558060026,
          "evaluation_seconds": 12.734534774906933,
          "tasks_per_second": 160.82252207874373,
          "input_tokens_per_second": 107982.27216824649,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "02ff3d0e5dbf",
          "lease": "lease-0006",
          "hours": 4.779901470555555,
          "research_hours": 4.862615659166667,
          "nll": 3.044046764266998,
          "score": 71.87922914237261,
          "ppl": 20.990013232823408,
          "seconds_per_task": 0.006155366227176273,
          "ms": 6.155366227176273,
          "evaluation_seconds": 12.606190033257008,
          "tasks_per_second": 162.45987047609714,
          "input_tokens_per_second": 109081.64928279446,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 27505,
      "output_tokens": 52761,
      "total_tokens": 20523440,
      "agent_turns": 190
    },
    {
      "model": "kimi-k3",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.161392606646895,
      "score": 70.795193383331,
      "seconds_per_task": 0.006348233416929361,
      "ms": 6.348233416929361,
      "ppl": 23.60344336409984,
      "evaluation_seconds": 13.001182037871331,
      "tasks_per_second": 157.52413850020338,
      "input_tokens_per_second": 105767.61374423032,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.371758672777778,
      "research_hours": 7.47693991,
      "leases": 4,
      "candidates": 3,
      "first_analysis": 3.2000519050285434,
      "final_analysis": 3.1733102544637486,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "ca16706d3790",
          "lease": "lease-0002",
          "hours": 5.675278240277778,
          "research_hours": 5.714804040555556,
          "nll": 3.2000519050285434,
          "score": 70.43806047589068,
          "ppl": 24.533803591836982,
          "seconds_per_task": 0.006408895285176186,
          "ms": 6.408895285176186,
          "evaluation_seconds": 13.125417544040829,
          "tasks_per_second": 156.0331313749198,
          "input_tokens_per_second": 104766.49564754772,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "3452f193a0c4",
          "lease": "lease-0003",
          "hours": 7.366771342222223,
          "research_hours": 7.435580319166666,
          "nll": 3.1733102544637486,
          "score": 70.68509867409274,
          "ppl": 23.88642377048062,
          "seconds_per_task": 0.006442496038516765,
          "ms": 6.442496038516765,
          "evaluation_seconds": 13.194231886882335,
          "tasks_per_second": 155.21934263078364,
          "input_tokens_per_second": 104220.08736765679,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "1dcbf468c8cc",
          "lease": "lease-0004",
          "hours": 7.371758672777778,
          "research_hours": 7.4716574994444445,
          "nll": 3.1733102544637486,
          "score": 70.68509867409274,
          "ppl": 23.88642377048062,
          "seconds_per_task": 0.006529467514610587,
          "ms": 6.529467514610587,
          "evaluation_seconds": 13.372349469922483,
          "tasks_per_second": 153.1518455007796,
          "input_tokens_per_second": 102831.8922634297,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 57309,
      "output_tokens": 27996,
      "total_tokens": 6859755,
      "agent_turns": 94
    },
    {
      "model": "glm-5.3",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.247955036400625,
      "score": 69.99553344362232,
      "seconds_per_task": 0.006678744517557789,
      "ms": 6.678744517557789,
      "ppl": 25.73765350011012,
      "evaluation_seconds": 13.678068771958351,
      "tasks_per_second": 149.7287397910033,
      "input_tokens_per_second": 100533.49072342178,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 5.712417235833334,
      "research_hours": 5.945129551666667,
      "leases": 2,
      "candidates": 1,
      "first_analysis": 3.2834213768823868,
      "final_analysis": 3.2834213768823868,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "854b869dfd03",
          "lease": "lease-0002",
          "hours": 5.712417235833334,
          "research_hours": 5.940207638333334,
          "nll": 3.2834213768823868,
          "score": 69.66789694159692,
          "ppl": 26.666854157368842,
          "seconds_per_task": 0.0067885132943956705,
          "ms": 6.7885132943956705,
          "evaluation_seconds": 13.902875226922333,
          "tasks_per_second": 147.30765878083506,
          "input_tokens_per_second": 98907.88614265695,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 95105,
      "output_tokens": 25630,
      "total_tokens": 8064366,
      "agent_turns": 74
    },
    {
      "model": "gpt-5.6-luna",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.7066179119344222,
      "score": 65.75842586196788,
      "seconds_per_task": 0.011390191824148133,
      "ms": 11.390191824148133,
      "ppl": 40.71586874584725,
      "evaluation_seconds": 23.327112855855376,
      "tasks_per_second": 87.79483396231473,
      "input_tokens_per_second": 58948.74382857169,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 1.9637719319444444,
      "research_hours": 1.9801401305555555,
      "leases": 1,
      "candidates": 1,
      "first_analysis": 3.793098750428555,
      "final_analysis": 3.793098750428555,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "f7e964c6a38c",
          "lease": "lease-0001",
          "hours": 1.9637719319444444,
          "research_hours": 1.976775925277778,
          "nll": 3.793098750428555,
          "score": 64.95951965874647,
          "ppl": 44.39375251555291,
          "seconds_per_task": 0.01147692753318097,
          "ms": 11.47692753318097,
          "evaluation_seconds": 23.504747587954625,
          "tasks_per_second": 87.1313334608847,
          "input_tokens_per_second": 58503.24471064278,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 1157,
      "output_tokens": 4944,
      "total_tokens": 951006,
      "agent_turns": 21
    },
    {
      "model": "gpt-5.6-terra",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.64037417043656,
      "score": 66.37038264833541,
      "seconds_per_task": 0.02178464821781745,
      "ms": 21.78464821781745,
      "ppl": 38.10609223140406,
      "evaluation_seconds": 44.614959550090134,
      "tasks_per_second": 45.903885617125084,
      "input_tokens_per_second": 30821.590199048424,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.567380592777778,
      "research_hours": 7.586737426666667,
      "leases": 2,
      "candidates": 1,
      "first_analysis": 3.715562529614639,
      "final_analysis": 3.715562529614639,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "0888849521bc",
          "lease": "lease-0002",
          "hours": 7.567380592777778,
          "research_hours": 7.580936770555555,
          "nll": 3.715562529614639,
          "score": 65.67579587508756,
          "ppl": 41.08169025306201,
          "seconds_per_task": 0.022030572419453165,
          "ms": 22.030572419453165,
          "evaluation_seconds": 45.11861231504008,
          "tasks_per_second": 45.391466956028445,
          "input_tokens_per_second": 30477.53309428835,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 2629,
      "output_tokens": 7189,
      "total_tokens": 656265,
      "agent_turns": 27
    },
    {
      "model": "gpt-5.6-sol",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.153931641503035,
      "score": 70.8641174529774,
      "seconds_per_task": 0.005515821313451852,
      "ms": 5.515821313451852,
      "ppl": 23.427994219739112,
      "evaluation_seconds": 11.296402049949393,
      "tasks_per_second": 181.29666339285214,
      "input_tokens_per_second": 121729.37842683816,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 6.843067181388888,
      "research_hours": 6.8567462425,
      "leases": 1,
      "candidates": 1,
      "first_analysis": 3.1705159381380508,
      "final_analysis": 3.1705159381380508,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "e5344a8a2605",
          "lease": "lease-0001",
          "hours": 6.843067181388888,
          "research_hours": 6.854362885555555,
          "nll": 3.1705159381380508,
          "score": 70.7109124460824,
          "ppl": 23.81977071480709,
          "seconds_per_task": 0.005449867834386168,
          "ms": 5.449867834386168,
          "evaluation_seconds": 11.161329324822873,
          "tasks_per_second": 183.49068828613755,
          "input_tokens_per_second": 123202.52901612349,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 2565,
      "output_tokens": 6781,
      "total_tokens": 891949,
      "agent_turns": 22
    },
    {
      "model": "claude-sonnet-5",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.4941142968437533,
      "score": 67.72152496298354,
      "seconds_per_task": 0.005543276047319523,
      "ms": 5.543276047319523,
      "ppl": 32.92111669940808,
      "evaluation_seconds": 11.352629344910383,
      "tasks_per_second": 180.39873740070271,
      "input_tokens_per_second": 121126.47724348433,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 6.537855405,
      "research_hours": 6.558599476944444,
      "leases": 2,
      "candidates": 1,
      "first_analysis": 3.551453179774088,
      "final_analysis": 3.551453179774088,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "693aaf436cbd",
          "lease": "lease-0002",
          "hours": 6.5378554047222215,
          "research_hours": 6.552582671666666,
          "nll": 3.551453179774088,
          "score": 67.19183087055242,
          "ppl": 34.86394427244461,
          "seconds_per_task": 0.005645032651386828,
          "ms": 5.645032651386828,
          "evaluation_seconds": 11.561026870040223,
          "tasks_per_second": 177.14689387214221,
          "input_tokens_per_second": 118943.0675542765,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 47015,
      "output_tokens": 30585,
      "total_tokens": 7086043,
      "agent_turns": 88
    },
    {
      "model": "claude-opus-5",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.916379053399159,
      "score": 73.05861787101352,
      "seconds_per_task": 0.005186905573623335,
      "ms": 5.186905573623335,
      "ppl": 18.474271851135924,
      "evaluation_seconds": 10.62278261478059,
      "tasks_per_second": 192.7931761636921,
      "input_tokens_per_second": 129448.56822040901,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 6.791675280833333,
      "research_hours": 6.829935034999999,
      "leases": 2,
      "candidates": 2,
      "first_analysis": 4.661999963745717,
      "final_analysis": 2.9361170224208113,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "fdd09e18ef04",
          "lease": "lease-0001",
          "hours": 0.7064001594444445,
          "research_hours": 0.7185653572222223,
          "nll": 4.661999963745717,
          "score": 56.93264825702211,
          "ppl": 105.84756189662814,
          "seconds_per_task": 0.005142097363773246,
          "ms": 5.142097363773246,
          "evaluation_seconds": 10.531015401007608,
          "tasks_per_second": 194.4731749042972,
          "input_tokens_per_second": 130576.58237480406,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "d13371377c0e",
          "lease": "lease-0002",
          "hours": 6.791675280833333,
          "research_hours": 6.826158202499999,
          "nll": 2.9361170224208113,
          "score": 72.87627937655662,
          "ppl": 18.842538922715907,
          "seconds_per_task": 0.004996623987267412,
          "ms": 4.996623987267412,
          "evaluation_seconds": 10.23308592592366,
          "tasks_per_second": 200.13513175060564,
          "input_tokens_per_second": 134378.23252479726,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 67332,
      "output_tokens": 62565,
      "total_tokens": 9089014,
      "agent_turns": 93
    },
    {
      "model": "deepseek-v4-pro",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.1074060893991597,
      "score": 71.29391846822321,
      "seconds_per_task": 0.006002029170872447,
      "ms": 6.002029170872447,
      "ppl": 22.362961579383267,
      "evaluation_seconds": 12.292155741946772,
      "tasks_per_second": 166.61031986531336,
      "input_tokens_per_second": 111868.41664456634,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.57606954,
      "research_hours": 7.794683673055555,
      "leases": 8,
      "candidates": 4,
      "first_analysis": 10.836687763656013,
      "final_analysis": 3.191800861126631,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "4496ac61c6b7",
          "lease": "lease-0004",
          "hours": 0.3996331513888889,
          "research_hours": 0.5588556827777778,
          "nll": 10.836687763656013,
          "score": 0.0,
          "ppl": 50852.66268652917,
          "seconds_per_task": 0.003965667229522296,
          "ms": 3.965667229522296,
          "evaluation_seconds": 8.121686486061662,
          "tasks_per_second": 252.1643754058658,
          "input_tokens_per_second": 169312.617811576,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "d2bd202f4559",
          "lease": "lease-0006",
          "hours": 3.447408520555556,
          "research_hours": 3.6452925708333335,
          "nll": 3.4145236583375556,
          "score": 68.45677980010296,
          "ppl": 30.40246401758441,
          "seconds_per_task": 0.006291344190003656,
          "ms": 6.291344190003656,
          "evaluation_seconds": 12.884672901127487,
          "tasks_per_second": 158.94854419011193,
          "input_tokens_per_second": 106724.01313964828,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "0c9f7d59f00e",
          "lease": "lease-0007",
          "hours": 6.7680912475,
          "research_hours": 6.974888850555556,
          "nll": 3.1799817861989528,
          "score": 70.62346735574462,
          "ppl": 24.046315573268107,
          "seconds_per_task": 0.006129045676289024,
          "ms": 6.129045676289024,
          "evaluation_seconds": 12.552285545039922,
          "tasks_per_second": 163.15753753779717,
          "input_tokens_per_second": 109550.0891105347,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "eb50bfc4d90a",
          "lease": "lease-0008",
          "hours": 7.576069539722223,
          "research_hours": 7.791797518055556,
          "nll": 3.191800861126631,
          "score": 70.51428325854485,
          "ppl": 24.332206936465926,
          "seconds_per_task": 0.00612039860538971,
          "ms": 6.12039860538971,
          "evaluation_seconds": 12.534576343838125,
          "tasks_per_second": 163.3880510853306,
          "input_tokens_per_second": 109704.86455060665,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 268880,
      "output_tokens": 29239,
      "total_tokens": 39382351,
      "agent_turns": 159
    },
    {
      "model": "deepseek-v4-flash",
      "track": "A",
      "track_config": "preview-a-8h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "A",
        "track_config": "preview-a-8h"
      },
      "challenge_manifest_sha256": "a1940190ba00d0290dd7d6e39e330864fc8ddd36e4ffd47c7c0dd1c1e23d9b84",
      "max_leases": null,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.0783356460307036,
      "score": 71.562469952481,
      "seconds_per_task": 0.005576860892574587,
      "ms": 5.576860892574587,
      "ppl": 21.722218832285126,
      "evaluation_seconds": 11.421411107992753,
      "tasks_per_second": 179.3123442134747,
      "input_tokens_per_second": 120397.03211783491,
      "task_count": 2048,
      "budget_hours": 8.0,
      "lease_hours": 7.683254876666666,
      "research_hours": 7.720291221388888,
      "leases": 5,
      "candidates": 3,
      "first_analysis": 3.1922240685001784,
      "final_analysis": 3.0933389509519023,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "9a7170c7a785",
          "lease": "lease-0002",
          "hours": 6.160158230277777,
          "research_hours": 6.172569628611111,
          "nll": 3.1922240685001784,
          "score": 70.51037368733968,
          "ppl": 24.342506685167187,
          "seconds_per_task": 0.005620221803269487,
          "ms": 5.620221803269487,
          "evaluation_seconds": 11.51021425309591,
          "tasks_per_second": 177.92892078000617,
          "input_tokens_per_second": 119468.14974622539,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "567c63cfc203",
          "lease": "lease-0003",
          "hours": 7.322455534166666,
          "research_hours": 7.342337705833334,
          "nll": 3.104102615758474,
          "score": 71.3244358132175,
          "ppl": 22.289208013830294,
          "seconds_per_task": 0.0054669981797133005,
          "ms": 5.4669981797133005,
          "evaluation_seconds": 11.19641227205284,
          "tasks_per_second": 182.915736776858,
          "input_tokens_per_second": 122816.4850121116,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "e8625f49cf3c",
          "lease": "lease-0004",
          "hours": 7.680716035833333,
          "research_hours": 7.710738369166666,
          "nll": 3.0933389509519023,
          "score": 71.42387007788355,
          "ppl": 22.050581004406418,
          "seconds_per_task": 0.005568542564446943,
          "ms": 5.568542564446943,
          "evaluation_seconds": 11.40437517198734,
          "tasks_per_second": 179.58020225698286,
          "input_tokens_per_second": 120576.88205292291,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 50692,
      "output_tokens": 18937,
      "total_tokens": 5574220,
      "agent_turns": 86
    },
    {
      "model": "gpt-6-astra",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.9798472947759014,
      "score": 72.47230103336082,
      "seconds_per_task": 0.005764507180629153,
      "ms": 5.764507180629153,
      "ppl": 19.684810441846935,
      "evaluation_seconds": 11.805710705928504,
      "tasks_per_second": 173.4753672196584,
      "input_tokens_per_second": 116477.86687754939,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 11.583489100555555,
      "research_hours": 11.71038553888889,
      "leases": 12,
      "candidates": 12,
      "first_analysis": 3.5047896327052794,
      "final_analysis": 3.0083542582488505,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "7490e13edb12",
          "lease": "lease-0001",
          "hours": 0.9729784102777778,
          "research_hours": 0.9817937838888888,
          "nll": 3.5047896327052794,
          "score": 67.62290667724821,
          "ppl": 33.274443261469635,
          "seconds_per_task": 0.0056132210200985355,
          "ms": 5.6132210200985355,
          "evaluation_seconds": 11.4958766491618,
          "tasks_per_second": 178.15083290314584,
          "input_tokens_per_second": 119617.14986740598,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "2fb9d0a5cb99",
          "lease": "lease-0002",
          "hours": 1.9324687005555556,
          "research_hours": 1.9506840449999998,
          "nll": 3.3595027244865596,
          "score": 68.96506078032934,
          "ppl": 28.774878278514546,
          "seconds_per_task": 0.005895656113352743,
          "ms": 5.895656113352743,
          "evaluation_seconds": 12.074303720146418,
          "tasks_per_second": 169.61640583736826,
          "input_tokens_per_second": 113886.81549442794,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "c3c16a364a4f",
          "lease": "lease-0003",
          "hours": 2.8925760422222226,
          "research_hours": 2.917775905555555,
          "nll": 3.257579779839019,
          "score": 69.90662048473759,
          "ppl": 25.986567758165307,
          "seconds_per_task": 0.005554823182706059,
          "ms": 5.554823182706059,
          "evaluation_seconds": 11.376277878182009,
          "tasks_per_second": 180.0237320088459,
          "input_tokens_per_second": 120874.68456068946,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "9f6be959921c",
          "lease": "lease-0004",
          "hours": 3.8526311205555555,
          "research_hours": 3.8888481061111113,
          "nll": 3.2209011296548034,
          "score": 70.24545625076621,
          "ppl": 25.05068392754073,
          "seconds_per_task": 0.00560117136228655,
          "ms": 5.60117136228655,
          "evaluation_seconds": 11.471198949962854,
          "tasks_per_second": 178.5340842690756,
          "input_tokens_per_second": 119874.47920641745,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "e9a45d944016",
          "lease": "lease-0005",
          "hours": 4.813551375277778,
          "research_hours": 4.857600954722222,
          "nll": 3.185727039147054,
          "score": 70.5703929603151,
          "ppl": 24.184865359386702,
          "seconds_per_task": 0.0056192266586094775,
          "ms": 5.6192266586094775,
          "evaluation_seconds": 11.50817619683221,
          "tasks_per_second": 177.96043134651876,
          "input_tokens_per_second": 119489.3071222282,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "fd698b9068af",
          "lease": "lease-0006",
          "hours": 5.774112215,
          "research_hours": 5.826728645555556,
          "nll": 3.170551045326791,
          "score": 70.71058812740846,
          "ppl": 23.8206069746726,
          "seconds_per_task": 0.005561238094287546,
          "ms": 5.561238094287546,
          "evaluation_seconds": 11.389415617100894,
          "tasks_per_second": 179.81607387520253,
          "input_tokens_per_second": 120735.2551025813,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 7,
          "candidate_id": "a827cdf34b35",
          "lease": "lease-0007",
          "hours": 6.732917225277778,
          "research_hours": 6.796826123055555,
          "nll": 3.13942773509017,
          "score": 70.99810390599922,
          "ppl": 23.0906491088255,
          "seconds_per_task": 0.005890129996487303,
          "ms": 5.890129996487303,
          "evaluation_seconds": 12.062986232805997,
          "tasks_per_second": 169.77553986013382,
          "input_tokens_per_second": 113993.6640448386,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 8,
          "candidate_id": "09700d1d14b9",
          "lease": "lease-0008",
          "hours": 7.695833071666667,
          "research_hours": 7.767009400277778,
          "nll": 3.095501247949304,
          "score": 71.4038948728033,
          "ppl": 22.098312495735904,
          "seconds_per_task": 0.00595213110500481,
          "ms": 5.95213110500481,
          "evaluation_seconds": 12.18996450304985,
          "tasks_per_second": 168.00705198834694,
          "input_tokens_per_second": 112806.2349694257,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 9,
          "candidate_id": "b03e3cf163d6",
          "lease": "lease-0009",
          "hours": 8.670947354166668,
          "research_hours": 8.749919761666666,
          "nll": 3.073820335285254,
          "score": 71.60418219564173,
          "ppl": 21.624357367904363,
          "seconds_per_task": 0.005795558878503471,
          "ms": 5.795558878503471,
          "evaluation_seconds": 11.869304583175108,
          "tasks_per_second": 172.5459133387702,
          "input_tokens_per_second": 115853.7966874005,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 10,
          "candidate_id": "c67ccb5fdf3d",
          "lease": "lease-0010",
          "hours": 9.634433482777778,
          "research_hours": 9.721296252777778,
          "nll": 3.0534608980746825,
          "score": 71.79226178604401,
          "ppl": 21.188549081260742,
          "seconds_per_task": 0.0060774398266403296,
          "ms": 6.07743982664033,
          "evaluation_seconds": 12.446596764959395,
          "tasks_per_second": 164.54297015274773,
          "input_tokens_per_second": 110480.32052193554,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 11,
          "candidate_id": "3592100221cf",
          "lease": "lease-0011",
          "hours": 10.608175547777778,
          "research_hours": 10.717140435833333,
          "nll": 3.0216443582836234,
          "score": 72.08618158893587,
          "ppl": 20.525014427953895,
          "seconds_per_task": 0.0053606049394829824,
          "ms": 5.3606049394829824,
          "evaluation_seconds": 10.978518916061148,
          "tasks_per_second": 186.5461102411415,
          "input_tokens_per_second": 125254.05389503643,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 12,
          "candidate_id": "3eac602d7a2e",
          "lease": "lease-0012",
          "hours": 11.583489100555555,
          "research_hours": 11.705513757222223,
          "nll": 3.0083542582488505,
          "score": 72.20895495173036,
          "ppl": 20.25403956296183,
          "seconds_per_task": 0.005480467770553332,
          "ms": 5.480467770553332,
          "evaluation_seconds": 11.223997994093224,
          "tasks_per_second": 182.4661765867908,
          "input_tokens_per_second": 122514.63344199334,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 7079,
      "output_tokens": 14590,
      "total_tokens": 6403698,
      "agent_turns": 90
    },
    {
      "model": "claude-fable-5-1",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.8687905812539505,
      "score": 73.49823809510762,
      "seconds_per_task": 0.005295427327723701,
      "ms": 5.295427327723701,
      "ppl": 17.615700552981316,
      "evaluation_seconds": 10.845035167178139,
      "tasks_per_second": 188.84217233321212,
      "input_tokens_per_second": 126795.71608598111,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 10.569369016666666,
      "research_hours": 10.664261931111112,
      "leases": 12,
      "candidates": 12,
      "first_analysis": 3.8086576881528513,
      "final_analysis": 2.894117040934942,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "27477d6cc81c",
          "lease": "lease-0001",
          "hours": 0.7952009336111111,
          "research_hours": 0.8127404938888888,
          "nll": 3.8086576881528513,
          "score": 64.81578687261813,
          "ppl": 45.08987355518571,
          "seconds_per_task": 0.006142569972212186,
          "ms": 6.142569972212186,
          "evaluation_seconds": 12.579983303090557,
          "tasks_per_second": 162.79830828526318,
          "input_tokens_per_second": 109308.8891192864,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "52c45a8c3ad5",
          "lease": "lease-0002",
          "hours": 1.688287758333333,
          "research_hours": 1.7140195144444446,
          "nll": 3.5593511099377015,
          "score": 67.11887013716698,
          "ppl": 35.140387497197786,
          "seconds_per_task": 0.006000811380886262,
          "ms": 6.000811380886262,
          "evaluation_seconds": 12.289661708055064,
          "tasks_per_second": 166.64413135616832,
          "input_tokens_per_second": 111891.11894745726,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "e95680128e53",
          "lease": "lease-0003",
          "hours": 2.6085974566666663,
          "research_hours": 2.640416853888889,
          "nll": 3.284571573625453,
          "score": 69.65727147439561,
          "ppl": 26.6975439324247,
          "seconds_per_task": 0.005885664631819054,
          "ms": 5.885664631819054,
          "evaluation_seconds": 12.053841165965423,
          "tasks_per_second": 169.90434599243125,
          "input_tokens_per_second": 114080.14931229306,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "8072ab436213",
          "lease": "lease-0004",
          "hours": 3.5185296066666667,
          "research_hours": 3.5563300749999995,
          "nll": 3.125570600187011,
          "score": 71.12611551210499,
          "ppl": 22.77288559976363,
          "seconds_per_task": 0.005214940570340332,
          "ms": 5.214940570340332,
          "evaluation_seconds": 10.680198288057,
          "tasks_per_second": 191.756739412802,
          "input_tokens_per_second": 128752.66571948324,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "ff3d5f5a7eeb",
          "lease": "lease-0005",
          "hours": 4.412784115555556,
          "research_hours": 4.45665199,
          "nll": 3.070586983103579,
          "score": 71.63405176166489,
          "ppl": 21.554551119740093,
          "seconds_per_task": 0.005104206349187734,
          "ms": 5.104206349187734,
          "evaluation_seconds": 10.45341460313648,
          "tasks_per_second": 195.91684418462756,
          "input_tokens_per_second": 131545.91606721588,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "4283d2b22a64",
          "lease": "lease-0006",
          "hours": 5.316528137222223,
          "research_hours": 5.366587822777778,
          "nll": 3.029124211793015,
          "score": 72.0170830293949,
          "ppl": 20.679114132294096,
          "seconds_per_task": 0.005099856678157266,
          "ms": 5.099856678157266,
          "evaluation_seconds": 10.444506476866081,
          "tasks_per_second": 196.08394178664065,
          "input_tokens_per_second": 131658.11166336754,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 7,
          "candidate_id": "a90e0748b6b3",
          "lease": "lease-0007",
          "hours": 6.2234843699999995,
          "research_hours": 6.279726836944445,
          "nll": 3.0005903650624366,
          "score": 72.2806774573834,
          "ppl": 20.0973982233515,
          "seconds_per_task": 0.005023196350634862,
          "ms": 5.023196350634862,
          "evaluation_seconds": 10.287506126100197,
          "tasks_per_second": 199.07643066224435,
          "input_tokens_per_second": 133667.3809127807,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 8,
          "candidate_id": "0e25a05bc6b3",
          "lease": "lease-0008",
          "hours": 7.128240450555555,
          "research_hours": 7.190635310277778,
          "nll": 2.977362054925696,
          "score": 72.49525957038931,
          "ppl": 19.635949706964524,
          "seconds_per_task": 0.005415361600626056,
          "ms": 5.415361600626056,
          "evaluation_seconds": 11.090660558082163,
          "tasks_per_second": 184.65987569221463,
          "input_tokens_per_second": 123987.56528509136,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 9,
          "candidate_id": "76c764267970",
          "lease": "lease-0009",
          "hours": 8.001568577777778,
          "research_hours": 8.070287231388889,
          "nll": 2.9608073673150512,
          "score": 72.6481910504123,
          "ppl": 19.313558598452026,
          "seconds_per_task": 0.0051048328379010854,
          "ms": 5.104832837901085,
          "evaluation_seconds": 10.454697652021423,
          "tasks_per_second": 195.8928003627171,
          "input_tokens_per_second": 131529.77214354186,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 10,
          "candidate_id": "3141bcc7ce32",
          "lease": "lease-0010",
          "hours": 8.911891596944445,
          "research_hours": 8.986859668333333,
          "nll": 2.9312602234692275,
          "score": 72.92114627282848,
          "ppl": 18.75124637366357,
          "seconds_per_task": 0.0053020091264670555,
          "ms": 5.3020091264670555,
          "evaluation_seconds": 10.85851469100453,
          "tasks_per_second": 188.6077477702006,
          "input_tokens_per_second": 126638.31464345407,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 11,
          "candidate_id": "0ce223b87528",
          "lease": "lease-0011",
          "hours": 9.821834835277778,
          "research_hours": 9.903271934444444,
          "nll": 2.9052824789774854,
          "score": 73.16112754014195,
          "ppl": 18.270403928360032,
          "seconds_per_task": 0.005241267212795719,
          "ms": 5.241267212795719,
          "evaluation_seconds": 10.734115251805633,
          "tasks_per_second": 190.7935541921349,
          "input_tokens_per_second": 128105.94704288158,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 12,
          "candidate_id": "4eae8dc3718d",
          "lease": "lease-0012",
          "hours": 10.569369016666666,
          "research_hours": 10.657171010555556,
          "nll": 2.894117040934942,
          "score": 73.26427336838779,
          "ppl": 18.067541496757926,
          "seconds_per_task": 0.005488741457497781,
          "ms": 5.488741457497781,
          "evaluation_seconds": 11.240942504955456,
          "tasks_per_second": 182.1911284660659,
          "input_tokens_per_second": 122329.95581943412,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 44988,
      "output_tokens": 59387,
      "total_tokens": 11145438,
      "agent_turns": 116
    },
    {
      "model": "gpt-5.6-luna",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 5.671991233589605,
      "score": 47.60239308457137,
      "seconds_per_task": 0.003143889860893978,
      "ms": 3.143889860893978,
      "ppl": 290.61263627569065,
      "evaluation_seconds": 6.438686435110867,
      "tasks_per_second": 318.07730049285055,
      "input_tokens_per_second": 213569.02744966836,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 0.05856560444444444,
      "research_hours": 0.13004843416666667,
      "leases": 1,
      "candidates": 1,
      "first_analysis": 5.751995881837667,
      "final_analysis": 5.751995881837667,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "f386200aa94a",
          "lease": "lease-0001",
          "hours": 0.05856560444444444,
          "research_hours": 0.1263203386111111,
          "nll": 5.751995881837667,
          "score": 46.86331364356594,
          "ppl": 314.8183739308777,
          "seconds_per_task": 0.003170872048031015,
          "ms": 3.170872048031015,
          "evaluation_seconds": 6.493945954367518,
          "tasks_per_second": 315.37065666871047,
          "input_tokens_per_second": 211751.6852869973,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 1879,
      "output_tokens": 4862,
      "total_tokens": 652747,
      "agent_turns": 25
    },
    {
      "model": "gpt-5.6-terra",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 4.064540093997045,
      "score": 62.451956400067665,
      "seconds_per_task": 0.0054828894208185375,
      "ms": 5.4828894208185375,
      "ppl": 58.23811830232459,
      "evaluation_seconds": 11.228957533836365,
      "tasks_per_second": 182.38558600197166,
      "input_tokens_per_second": 122460.52190119885,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 1.6023312075,
      "research_hours": 1.6276828983333334,
      "leases": 2,
      "candidates": 1,
      "first_analysis": 4.078436688207989,
      "final_analysis": 4.078436688207989,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "3ceb187a21d1",
          "lease": "lease-0002",
          "hours": 1.6023312075,
          "research_hours": 1.6231616111111111,
          "nll": 4.078436688207989,
          "score": 62.323580270602555,
          "ppl": 59.05307927587528,
          "seconds_per_task": 0.00535395826182139,
          "ms": 5.35395826182139,
          "evaluation_seconds": 10.964906520210207,
          "tasks_per_second": 186.77769812493923,
          "input_tokens_per_second": 125409.55068476389,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 1670,
      "output_tokens": 5090,
      "total_tokens": 721410,
      "agent_turns": 27
    },
    {
      "model": "gpt-5.6-sol",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.1067920314864157,
      "score": 71.29959110836141,
      "seconds_per_task": 0.00634530802426525,
      "ms": 6.34530802426525,
      "ppl": 22.34923364117896,
      "evaluation_seconds": 12.995190833695233,
      "tasks_per_second": 157.59676223374422,
      "input_tokens_per_second": 105816.37604231964,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 10.3661413825,
      "research_hours": 10.432393133333333,
      "leases": 12,
      "candidates": 1,
      "first_analysis": 3.156662634956278,
      "final_analysis": 3.156662634956278,
      "reasoning_effort": "medium",
      "points": [
        {
          "index": 1,
          "candidate_id": "31f2ac5e86db",
          "lease": "lease-0012",
          "hours": 10.366141382499999,
          "research_hours": 10.426934074722222,
          "nll": 3.156662634956278,
          "score": 70.83888865491363,
          "ppl": 23.492063364963208,
          "seconds_per_task": 0.00659322065212109,
          "ms": 6.59322065212109,
          "evaluation_seconds": 13.502915895543993,
          "tasks_per_second": 151.670943953361,
          "input_tokens_per_second": 101837.5594306848,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 4322,
      "output_tokens": 9430,
      "total_tokens": 2030271,
      "agent_turns": 59
    },
    {
      "model": "claude-sonnet-5",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.101838232154234,
      "score": 71.34535408990632,
      "seconds_per_task": 0.0071573377445020014,
      "ms": 7.157337744502001,
      "ppl": 22.238793796756852,
      "evaluation_seconds": 14.658227700740099,
      "tasks_per_second": 139.71675442704412,
      "input_tokens_per_second": 93811.06830060844,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 7.2587802508333334,
      "research_hours": 7.371615345833334,
      "leases": 12,
      "candidates": 2,
      "first_analysis": 3.1226833883157874,
      "final_analysis": 3.1027547741720554,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "b5aed4c14167",
          "lease": "lease-0010",
          "hours": 6.196295706388889,
          "research_hours": 6.274013622222222,
          "nll": 3.1226833883157874,
          "score": 71.15278744908086,
          "ppl": 22.707230280135608,
          "seconds_per_task": 0.007350381064497924,
          "ms": 7.350381064497924,
          "evaluation_seconds": 15.053580420091748,
          "tasks_per_second": 136.04736832352324,
          "input_tokens_per_second": 91347.30486872565,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "f439a631afa8",
          "lease": "lease-0012",
          "hours": 7.2587802508333334,
          "research_hours": 7.368029894722222,
          "nll": 3.1027547741720554,
          "score": 71.33688711483646,
          "ppl": 22.25918592939291,
          "seconds_per_task": 0.007167745921833557,
          "ms": 7.167745921833557,
          "evaluation_seconds": 14.679543647915125,
          "tasks_per_second": 139.51387380430378,
          "input_tokens_per_second": 93674.84664247722,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 51449,
      "output_tokens": 33550,
      "total_tokens": 8792481,
      "agent_turns": 97
    },
    {
      "model": "claude-opus-5",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.843612031836784,
      "score": 73.73083643328002,
      "seconds_per_task": 0.006220763531018747,
      "ms": 6.220763531018747,
      "ppl": 17.177700014491887,
      "evaluation_seconds": 12.740123711526394,
      "tasks_per_second": 160.75197120316105,
      "input_tokens_per_second": 107934.90166472243,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 10.185028385277777,
      "research_hours": 10.254238075277778,
      "leases": 12,
      "candidates": 3,
      "first_analysis": 3.00897342493163,
      "final_analysis": 2.8259236885421797,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "2618f24b8925",
          "lease": "lease-0006",
          "hours": 5.819389105,
          "research_hours": 5.849830659999999,
          "nll": 3.00897342493163,
          "score": 72.20323511699803,
          "ppl": 20.266584072621367,
          "seconds_per_task": 0.006662470521860087,
          "ms": 6.662470521860087,
          "evaluation_seconds": 13.644739628769457,
          "tasks_per_second": 150.0944727213309,
          "input_tokens_per_second": 100779.05752782861,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "4296490ca675",
          "lease": "lease-0011",
          "hours": 10.100326319444443,
          "research_hours": 10.157837314722224,
          "nll": 2.8259236885421797,
          "score": 73.89424057492381,
          "ppl": 16.876526442881715,
          "seconds_per_task": 0.006083168002987804,
          "ms": 6.083168002987804,
          "evaluation_seconds": 12.458328070119023,
          "tasks_per_second": 164.38802931446915,
          "input_tokens_per_second": 110376.28743283388,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "ee3df937da65",
          "lease": "lease-0012",
          "hours": 10.185028384999999,
          "research_hours": 10.251176719999998,
          "nll": 2.8259236885421797,
          "score": 73.89424057492381,
          "ppl": 16.876526442881715,
          "seconds_per_task": 0.006379253887189407,
          "ms": 6.379253887189407,
          "evaluation_seconds": 13.064711960963905,
          "tasks_per_second": 156.75814408455585,
          "input_tokens_per_second": 105253.29636877397,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 63112,
      "output_tokens": 49984,
      "total_tokens": 9082309,
      "agent_turns": 101
    },
    {
      "model": "deepseek-v4-flash",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 2.994324602608485,
      "score": 72.3385602968602,
      "seconds_per_task": 0.005876307781818468,
      "ms": 5.876307781818468,
      "ppl": 19.97186638717022,
      "evaluation_seconds": 12.034678337164223,
      "tasks_per_second": 170.1748848305802,
      "input_tokens_per_second": 114261.79923343268,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 10.24964059611111,
      "research_hours": 10.357881725,
      "leases": 12,
      "candidates": 11,
      "first_analysis": 3.614510028096719,
      "final_analysis": 3.0051731386598104,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "c911d107882f",
          "lease": "lease-0002",
          "hours": 0.9746226380555556,
          "research_hours": 0.9914455191666667,
          "nll": 3.614510028096719,
          "score": 66.60931446393873,
          "ppl": 37.13314726875324,
          "seconds_per_task": 0.006135101407835464,
          "ms": 6.135101407835464,
          "evaluation_seconds": 12.56468768324703,
          "tasks_per_second": 162.99649077077143,
          "input_tokens_per_second": 109441.95627189985,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "456d034ad512",
          "lease": "lease-0003",
          "hours": 1.8544543666666666,
          "research_hours": 1.8809778636111112,
          "nll": 3.411208341851638,
          "score": 68.4874065487835,
          "ppl": 30.30183712458159,
          "seconds_per_task": 0.006025114841577306,
          "ms": 6.025114841577306,
          "evaluation_seconds": 12.339435195550323,
          "tasks_per_second": 165.9719401693946,
          "input_tokens_per_second": 111439.78457748788,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 3,
          "candidate_id": "78040e69127c",
          "lease": "lease-0004",
          "hours": 2.787246035,
          "research_hours": 2.8265486016666665,
          "nll": 3.3060687789490206,
          "score": 69.45868123193297,
          "ppl": 27.277679824835673,
          "seconds_per_task": 0.005814222734898067,
          "ms": 5.814222734898067,
          "evaluation_seconds": 11.907528161071241,
          "tasks_per_second": 171.9920349796389,
          "input_tokens_per_second": 115481.90198664129,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 4,
          "candidate_id": "ebb548cd6a05",
          "lease": "lease-0005",
          "hours": 3.720069257222222,
          "research_hours": 3.767063778333333,
          "nll": 3.266834234304885,
          "score": 69.8211282391841,
          "ppl": 26.228175515832582,
          "seconds_per_task": 0.005848404427979403,
          "ms": 5.848404427979403,
          "evaluation_seconds": 11.977532268501818,
          "tasks_per_second": 170.98680713937824,
          "input_tokens_per_second": 114806.95431864628,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 5,
          "candidate_id": "9d2f0674fbc2",
          "lease": "lease-0006",
          "hours": 4.652918748888888,
          "research_hours": 4.70792152,
          "nll": 3.237184261454949,
          "score": 70.09503339144241,
          "ppl": 25.46192658846197,
          "seconds_per_task": 0.0058257962909920025,
          "ms": 5.8257962909920025,
          "evaluation_seconds": 11.931230803951621,
          "tasks_per_second": 171.65035474141553,
          "input_tokens_per_second": 115252.48506168918,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 6,
          "candidate_id": "191a0ee5dee1",
          "lease": "lease-0007",
          "hours": 5.585693151388889,
          "research_hours": 5.649783705277778,
          "nll": 3.204359583612304,
          "score": 70.3982663295575,
          "ppl": 24.63971528513866,
          "seconds_per_task": 0.0057738114046514966,
          "ms": 5.773811404651497,
          "evaluation_seconds": 11.824765756726265,
          "tasks_per_second": 173.19581986941594,
          "input_tokens_per_second": 116290.16830357096,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 7,
          "candidate_id": "ded9441862e4",
          "lease": "lease-0008",
          "hours": 6.518581679722223,
          "research_hours": 6.590719441388889,
          "nll": 3.155327833012518,
          "score": 70.85121949688407,
          "ppl": 23.46072703167266,
          "seconds_per_task": 0.005811055315916747,
          "ms": 5.811055315916747,
          "evaluation_seconds": 11.901041286997497,
          "tasks_per_second": 172.0857822951632,
          "input_tokens_per_second": 115544.84744980863,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 8,
          "candidate_id": "eee049515983",
          "lease": "lease-0009",
          "hours": 7.451308951111111,
          "research_hours": 7.531788938055556,
          "nll": 3.1091322860995847,
          "score": 71.27797193861083,
          "ppl": 22.401597787126622,
          "seconds_per_task": 0.005890916801490675,
          "ms": 5.890916801490675,
          "evaluation_seconds": 12.064597609452903,
          "tasks_per_second": 169.7528642310741,
          "input_tokens_per_second": 113978.43877715182,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 9,
          "candidate_id": "b3902e093411",
          "lease": "lease-0010",
          "hours": 8.38405085,
          "research_hours": 8.473508575555556,
          "nll": 3.0596848605721276,
          "score": 71.73476509273704,
          "ppl": 21.320837066804042,
          "seconds_per_task": 0.005902429093112005,
          "ms": 5.902429093112005,
          "evaluation_seconds": 12.088174782693386,
          "tasks_per_second": 169.42177266762533,
          "input_tokens_per_second": 113756.13148551868,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 10,
          "candidate_id": "4319da58e709",
          "lease": "lease-0011",
          "hours": 9.316849079166667,
          "research_hours": 9.414204247222221,
          "nll": 3.007795899160788,
          "score": 72.21411304856252,
          "ppl": 20.24273369256334,
          "seconds_per_task": 0.005729739699290803,
          "ms": 5.729739699290803,
          "evaluation_seconds": 11.734506904147565,
          "tasks_per_second": 174.52799821321284,
          "input_tokens_per_second": 117184.6428002841,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 11,
          "candidate_id": "fde73befb054",
          "lease": "lease-0012",
          "hours": 10.24964059611111,
          "research_hours": 10.354951192777778,
          "nll": 3.0051731386598104,
          "score": 72.23834199534751,
          "ppl": 20.189711412967437,
          "seconds_per_task": 0.005748589970153262,
          "ms": 5.748589970153262,
          "evaluation_seconds": 11.77311225887388,
          "tasks_per_second": 173.9557013445054,
          "input_tokens_per_second": 116800.38122150135,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 26922,
      "output_tokens": 14237,
      "total_tokens": 2392353,
      "agent_turns": 50
    },
    {
      "model": "deepseek-v4-pro",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.245675275360559,
      "score": 70.01659377639066,
      "seconds_per_task": 0.08424569574481211,
      "ms": 84.24569574481211,
      "ppl": 25.67904463288794,
      "evaluation_seconds": 172.5351848853752,
      "tasks_per_second": 11.870042631365893,
      "input_tokens_per_second": 7969.991749297737,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 11.784627684444445,
      "research_hours": 11.856942313055555,
      "leases": 12,
      "candidates": 1,
      "first_analysis": 3.2617016712297926,
      "final_analysis": 3.2617016712297926,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "b788c1bd7f79",
          "lease": "lease-0012",
          "hours": 11.784627684444445,
          "research_hours": 11.853143955277778,
          "nll": 3.2617016712297926,
          "score": 69.86854263236665,
          "ppl": 26.093902627452128,
          "seconds_per_task": 0.08369561723202423,
          "ms": 83.69561723202423,
          "evaluation_seconds": 171.40862409118563,
          "tasks_per_second": 11.9480569362164,
          "input_tokens_per_second": 8022.373479110799,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 52369,
      "output_tokens": 20308,
      "total_tokens": 7910151,
      "agent_turns": 89
    },
    {
      "model": "kimi-k3",
      "track": "B",
      "track_config": "preview-b-12x1h",
      "comparison_key": {
        "challenge_id": "lm-pretrain",
        "runtime_profile": "h200-sxm-141gb-x1",
        "track": "B",
        "track_config": "preview-b-12x1h"
      },
      "challenge_manifest_sha256": "051cc6cf3861ea4939ed2ff4632449af93aa764c7e2e4b83b870aaad04b945bc",
      "max_leases": 12,
      "status": "finalized",
      "outcome": "success",
      "test_nll": 3.014810705797679,
      "score": 72.14931057168793,
      "seconds_per_task": 0.005191842980366346,
      "ms": 5.191842980366346,
      "ppl": 20.385231769114487,
      "evaluation_seconds": 10.632894423790276,
      "tasks_per_second": 192.60983118742897,
      "input_tokens_per_second": 129325.46352790935,
      "task_count": 2048,
      "budget_hours": 12.0,
      "lease_hours": 10.914635439166666,
      "research_hours": 11.099531710555556,
      "leases": 12,
      "candidates": 2,
      "first_analysis": 3.0477444395182536,
      "final_analysis": 3.0139275311561993,
      "reasoning_effort": "provider-default",
      "points": [
        {
          "index": 1,
          "candidate_id": "7378e66b05ef",
          "lease": "lease-0010",
          "hours": 9.229540942500002,
          "research_hours": 9.355648016111111,
          "nll": 3.0477444395182536,
          "score": 71.84507018014277,
          "ppl": 21.06777115845918,
          "seconds_per_task": 0.005235196976173029,
          "ms": 5.235196976173029,
          "evaluation_seconds": 10.721683407202363,
          "tasks_per_second": 191.01478025589174,
          "input_tokens_per_second": 128254.48651806531,
          "task_count": 2048,
          "final": false
        },
        {
          "index": 2,
          "candidate_id": "e56dbe2e1360",
          "lease": "lease-0012",
          "hours": 10.91463543861111,
          "research_hours": 11.092308199166666,
          "nll": 3.0139275311561993,
          "score": 72.15746930039468,
          "ppl": 20.367235997229727,
          "seconds_per_task": 0.005304723154040403,
          "ms": 5.304723154040403,
          "evaluation_seconds": 10.864073019474745,
          "tasks_per_second": 188.51125138139182,
          "input_tokens_per_second": 126573.52334939326,
          "task_count": 2048,
          "final": true
        }
      ],
      "reasoning_tokens": 64136,
      "output_tokens": 19973,
      "total_tokens": 5241646,
      "agent_turns": 71
    }
  ]
}
