{
  "collectedAt": "2026-09-26",
  "supplement": true,
  "supplementNote": "GPT-6 Luna 榜分启用：chatgpt_plus × gpt-6-luna 采用行新增（本机 Codex 周窗 48%→35% 实测 143,157,917 tok，chatgpt-gpt6luna-plus-round1-2026-09-26.json），沿 scores-gpt6sol-round1 先例把 scores-new-models-round1-2026-09-23.json 中 gpt-6-luna 的 11 条 unmappedOfficialRows 原样移入 scores[]；原文件只追加不改。无 baseSnapshot → 追加到各板现行快照。AA 独立实测 TB4 行保留 boardId=terminal_bench_4 + agentHarness=Artificial Analysis，由 benchmark_configs 路由到 aa_terminal_bench_4。",
  "sourceArchive": "scores-new-models-round1-2026-09-23.json",
  "boards": [
    {
      "boardId": "aa_intelligence_index",
      "name": "Artificial Analysis Intelligence Index v4.3",
      "url": "https://artificialanalysis.ai/leaderboards/models",
      "metric": "Intelligence Index",
      "snapshotDate": "2026-09-22",
      "note": "沿用 round4 元数据；线上版本已是 v4.3.2，本档读数为 v4.3.2 原始精度。"
    },
    {
      "boardId": "aa_coding_agent_index",
      "name": "Artificial Analysis Coding Agent Index v1.5",
      "url": "https://artificialanalysis.ai/agents/coding-agents",
      "metric": "Coding Agent Index",
      "snapshotDate": "2026-09-22",
      "note": "沿用 round4 元数据；v1.5 板面 2026-09-23 为 19 配置。"
    },
    {
      "boardId": "terminal_bench_4",
      "name": "Terminal-Bench 4.0",
      "url": "https://www.tbench.ai/leaderboard",
      "metric": "Resolution Rate %",
      "snapshotDate": "2026-09-03",
      "note": "官方榜 2026-09-23 复核：27 行无新模型行（榜首 Astra Codex max 58.18）。本榜现只收官方/厂商自报行；AA 独立实测行已路由至 aa_terminal_bench_4。"
    }
  ],
  "scores": [
    {
      "boardId": "aa_intelligence_index",
      "model": "gpt-6-luna",
      "variantLabel": "GPT-6 Luna (max)",
      "score": 37.2559686869738,
      "secondary": {
        "scoreRoundedDisplay": 37,
        "intelligenceIndexIsEstimated": false,
        "slug": "gpt-6-luna",
        "shortName": "GPT-6 Luna (max)",
        "releaseDate": "2026-09-22",
        "isReasoning": true,
        "deprecated": false,
        "modelCreatorName": "OpenAI",
        "effort": "max",
        "contextWindowTokens": 1000000,
        "costPerTaskUsd": 0.06809498628701058,
        "terminalbenchV40": 0.126262626262626,
        "scicode": 0.546296296296296,
        "hle": 0.385078776645042,
        "price1mInputTokens": 0.1,
        "price1mOutputTokens": 0.5
      },
      "source": "https://artificialanalysis.ai/leaderboards/models",
      "checkedAt": "2026-09-23",
      "note": "9/22 新发，无 adopted 套餐行——存档待采纳；AA v4.3.2 原始精度。"
    },
    {
      "boardId": "terminal_bench_4",
      "model": "gpt-6-luna",
      "variantLabel": "Artificial Analysis - GPT-6 Luna (max)",
      "score": 12.63,
      "secondary": {
        "agentHarness": "Artificial Analysis",
        "reasoningEffort": "max"
      },
      "source": "https://artificialanalysis.ai/models/claude-opus-5-5（AA 模型页 evals 载荷，页面注明 Independently benchmarked by Artificial Analysis）",
      "checkedAt": "2026-09-23",
      "note": "AA 独立实测 TB4；无 adopted 套餐行，存档待采纳。"
    },
    {
      "boardId": "aa_intelligence_index",
      "model": "gpt-6-luna",
      "variantLabel": "GPT-6 Luna (xhigh)",
      "score": 33.8845820217785,
      "secondary": {
        "scoreRoundedDisplay": 34,
        "intelligenceIndexIsEstimated": false,
        "slug": "gpt-6-luna-xhigh",
        "shortName": "GPT-6 Luna (xhigh)",
        "releaseDate": "2026-09-22",
        "isReasoning": true,
        "deprecated": false,
        "modelCreatorName": "OpenAI",
        "effort": "xhigh",
        "contextWindowTokens": 1000000,
        "costPerTaskUsd": 0.041708771594451556,
        "terminalbenchV40": 0.0808080808080808,
        "scicode": 0.517361111111111,
        "hle": 0.343373493975904,
        "price1mInputTokens": 0.1,
        "price1mOutputTokens": 0.5
      },
      "source": "https://artificialanalysis.ai/leaderboards/models",
      "checkedAt": "2026-09-23",
      "note": "9/22 新发，无 adopted 套餐行——存档待采纳；AA v4.3.2 原始精度。"
    },
    {
      "boardId": "terminal_bench_4",
      "model": "gpt-6-luna",
      "variantLabel": "Artificial Analysis - GPT-6 Luna (xhigh)",
      "score": 8.08,
      "secondary": {
        "agentHarness": "Artificial Analysis",
        "reasoningEffort": "xhigh"
      },
      "source": "https://artificialanalysis.ai/models/claude-opus-5-5（AA 模型页 evals 载荷，页面注明 Independently benchmarked by Artificial Analysis）",
      "checkedAt": "2026-09-23",
      "note": "AA 独立实测 TB4；无 adopted 套餐行，存档待采纳。"
    },
    {
      "boardId": "aa_intelligence_index",
      "model": "gpt-6-luna",
      "variantLabel": "GPT-6 Luna (high)",
      "score": 32.1482304150837,
      "secondary": {
        "scoreRoundedDisplay": 32,
        "intelligenceIndexIsEstimated": false,
        "slug": "gpt-6-luna-high",
        "shortName": "GPT-6 Luna (high)",
        "releaseDate": "2026-09-22",
        "isReasoning": true,
        "deprecated": false,
        "modelCreatorName": "OpenAI",
        "effort": "high",
        "contextWindowTokens": 1000000,
        "costPerTaskUsd": 0.02861964876935845,
        "terminalbenchV40": 0.0454545454545455,
        "scicode": 0.503472222222222,
        "hle": 0.329471733086191,
        "price1mInputTokens": 0.1,
        "price1mOutputTokens": 0.5
      },
      "source": "https://artificialanalysis.ai/leaderboards/models",
      "checkedAt": "2026-09-23",
      "note": "9/22 新发，无 adopted 套餐行——存档待采纳；AA v4.3.2 原始精度。"
    },
    {
      "boardId": "terminal_bench_4",
      "model": "gpt-6-luna",
      "variantLabel": "Artificial Analysis - GPT-6 Luna (high)",
      "score": 4.55,
      "secondary": {
        "agentHarness": "Artificial Analysis",
        "reasoningEffort": "high"
      },
      "source": "https://artificialanalysis.ai/models/claude-opus-5-5（AA 模型页 evals 载荷，页面注明 Independently benchmarked by Artificial Analysis）",
      "checkedAt": "2026-09-23",
      "note": "AA 独立实测 TB4；无 adopted 套餐行，存档待采纳。"
    },
    {
      "boardId": "aa_intelligence_index",
      "model": "gpt-6-luna",
      "variantLabel": "GPT-6 Luna (medium)",
      "score": 29.4619523515762,
      "secondary": {
        "scoreRoundedDisplay": 29,
        "intelligenceIndexIsEstimated": false,
        "slug": "gpt-6-luna-medium",
        "shortName": "GPT-6 Luna (medium)",
        "releaseDate": "2026-09-22",
        "isReasoning": true,
        "deprecated": false,
        "modelCreatorName": "OpenAI",
        "effort": "medium",
        "contextWindowTokens": 1000000,
        "costPerTaskUsd": 0.01725359419426588,
        "terminalbenchV40": 0.0252525252525253,
        "scicode": 0.509259259259259,
        "hle": 0.282669138090825,
        "price1mInputTokens": 0.1,
        "price1mOutputTokens": 0.5
      },
      "source": "https://artificialanalysis.ai/leaderboards/models",
      "checkedAt": "2026-09-23",
      "note": "9/22 新发，无 adopted 套餐行——存档待采纳；AA v4.3.2 原始精度。"
    },
    {
      "boardId": "terminal_bench_4",
      "model": "gpt-6-luna",
      "variantLabel": "Artificial Analysis - GPT-6 Luna (medium)",
      "score": 2.53,
      "secondary": {
        "agentHarness": "Artificial Analysis",
        "reasoningEffort": "medium"
      },
      "source": "https://artificialanalysis.ai/models/claude-opus-5-5（AA 模型页 evals 载荷，页面注明 Independently benchmarked by Artificial Analysis）",
      "checkedAt": "2026-09-23",
      "note": "AA 独立实测 TB4；无 adopted 套餐行，存档待采纳。"
    },
    {
      "boardId": "aa_intelligence_index",
      "model": "gpt-6-luna",
      "variantLabel": "GPT-6 Luna (low)",
      "score": 20.9225480080866,
      "secondary": {
        "scoreRoundedDisplay": 21,
        "intelligenceIndexIsEstimated": false,
        "slug": "gpt-6-luna-low",
        "shortName": "GPT-6 Luna (low)",
        "releaseDate": "2026-09-22",
        "isReasoning": true,
        "deprecated": false,
        "modelCreatorName": "OpenAI",
        "effort": "low",
        "contextWindowTokens": 1000000,
        "costPerTaskUsd": 0.004483809259539013,
        "terminalbenchV40": 0,
        "scicode": 0.46875,
        "hle": 0.202502316960148,
        "price1mInputTokens": 0.1,
        "price1mOutputTokens": 0.5
      },
      "source": "https://artificialanalysis.ai/leaderboards/models",
      "checkedAt": "2026-09-23",
      "note": "9/22 新发，无 adopted 套餐行——存档待采纳；AA v4.3.2 原始精度。"
    },
    {
      "boardId": "terminal_bench_4",
      "model": "gpt-6-luna",
      "variantLabel": "Artificial Analysis - GPT-6 Luna (low)",
      "score": 0,
      "secondary": {
        "agentHarness": "Artificial Analysis",
        "reasoningEffort": "low"
      },
      "source": "https://artificialanalysis.ai/models/claude-opus-5-5（AA 模型页 evals 载荷，页面注明 Independently benchmarked by Artificial Analysis）",
      "checkedAt": "2026-09-23",
      "note": "AA 独立实测 TB4；无 adopted 套餐行，存档待采纳。"
    },
    {
      "boardId": "aa_coding_agent_index",
      "model": "gpt-6-luna",
      "variantLabel": "Codex - GPT-6 Luna (max) ({'reasoning_effort': 'max'})",
      "score": 41.0744,
      "secondary": {
        "configurationRank": 18,
        "agentHarness": "Codex",
        "provider": "openai",
        "hostModelSlug": "openai_gpt-6-luna",
        "displayModel": "GPT-6 Luna (max)",
        "isUnavailable": false,
        "variantOf": null,
        "deepSWE": 63.716814159292,
        "terminalBenchV2_1": null,
        "terminalBenchV4": 15.151515151515198,
        "sweAtlasQnA": 44.3548387096774,
        "meanCostUsdPerTask": 0.17591172304455457,
        "agentWallTimeSec": 1285.9268514851478,
        "meanSteps": 88.64301430143014,
        "meanTotalTokens": 10218069.248074815,
        "cacheHitRate": 0.9606438143896164,
        "hostConfigId": "ac8276cd5d15ac2daacb9956ad44c132",
        "versions": {
          "deep-swe-v1.1": {
            "min": {
              "version": "0.154.0",
              "dateReleased": null
            },
            "max": {
              "version": "0.154.0",
              "dateReleased": null
            }
          },
          "swe-atlas-qna": {
            "min": {
              "version": "0.154.0",
              "dateReleased": null
            },
            "max": {
              "version": "0.154.0",
              "dateReleased": null
            }
          },
          "terminal-bench-v4": {
            "min": {
              "version": "0.154.0",
              "dateReleased": null
            },
            "max": {
              "version": "0.154.0",
              "dateReleased": null
            }
          }
        }
      },
      "source": "https://artificialanalysis.ai/agents/coding-agents",
      "checkedAt": "2026-09-23",
      "note": "AA Coding Agent Index v1.5（19 配置板面）原始精度 indexScore×100。无 adopted 套餐行，存档待采纳；TB4 子项为 Codex harness 运行，与 tbench.ai 独立榜不互灌。"
    }
  ]
}