{
  "schemaVersion": "2.0",
  "meta": {
    "date": "2026-11-18",
    "comparedTo": null,
    "generatedAt": "2026-11-18T12:00:00Z"
  },
  "provenance": {
    "briefSources": [
      "aa_models",
      "cursorbench",
      "deepmind_eval"
    ],
    "scatterSources": {
      "aa": "aa_index",
      "cursor": "cursorbench"
    },
    "tokenBreakdownSource": "aa_index",
    "tableSources": [
      "aa_index",
      "aa_models",
      "cursorbench",
      "terminal_bench",
      "frontier_code",
      "aa_omniscience"
    ]
  },
  "models": [
    {
      "id": "gemini-4-neon-high",
      "name": "Gemini 4 Neon",
      "shortLabel": "Neon high",
      "effort": "high",
      "vendor": "google",
      "weights": "closed",
      "aaSlug": "gemini-4-neon-high",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 3,
        "outputPer1M": 15
      },
      "metrics": {
        "aaIndex": 60,
        "aaCostUsd": 3.74,
        "aaReasoningTokensK": 35,
        "aaAnswerTokensK": 27,
        "aaTotalOutputTokensK": 62,
        "aaTimeMin": 14.8,
        "cursorBenchPct": 63.7,
        "cursorCostUsd": 6.54,
        "terminalBenchPct": null,
        "terminalBenchNote": "xhigh 65.2%",
        "frontierCodePct": 46.2,
        "automationBenchPct": 56.3,
        "hallucinationPct": 13.9
      },
      "chartOffset": {
        "aa": {
          "dx": -8,
          "dy": -8,
          "anchor": "end"
        },
        "cursor": {
          "dx": -8,
          "dy": -8,
          "anchor": "end"
        }
      }
    },
    {
      "id": "gemini-4-neon-xhigh",
      "name": "Gemini 4 Neon",
      "shortLabel": "Neon xhigh",
      "effort": "xhigh",
      "vendor": "google",
      "weights": "closed",
      "aaSlug": "gemini-4-neon-xhigh",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 3,
        "outputPer1M": 15
      },
      "metrics": {
        "aaIndex": 61,
        "aaCostUsd": 6.44,
        "aaReasoningTokensK": 81,
        "aaAnswerTokensK": 61,
        "aaTotalOutputTokensK": 142,
        "aaTimeMin": 21.5,
        "cursorBenchPct": 65.4,
        "cursorCostUsd": 10.67,
        "terminalBenchPct": 65.2,
        "frontierCodePct": null,
        "hallucinationPct": 14.4
      },
      "chartOffset": {
        "aa": {
          "dx": 7,
          "dy": -6,
          "anchor": "start"
        },
        "cursor": {
          "dx": 7,
          "dy": -6,
          "anchor": "start"
        }
      }
    },
    {
      "id": "gpt-6-2-astra-max",
      "name": "GPT-6.2 Astra",
      "shortLabel": "Astra max",
      "effort": "max",
      "vendor": "openai",
      "weights": "closed",
      "aaSlug": "gpt-6-2-astra-max",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 8,
        "outputPer1M": 40,
        "contextCapK": 272
      },
      "metrics": {
        "aaIndex": 62,
        "aaCostUsd": 4,
        "aaReasoningTokensK": 26,
        "aaAnswerTokensK": 15,
        "aaTotalOutputTokensK": 41,
        "aaTimeMin": 13.7,
        "cursorBenchPct": null,
        "cursorCostUsd": null,
        "terminalBenchPct": 62.1,
        "frontierCodePct": null,
        "frontierCodeNote": "6.0: 44.0%",
        "hallucinationPct": 59.1
      },
      "chartOffset": {
        "aa": {
          "dx": 7,
          "dy": -6,
          "anchor": "start"
        }
      }
    },
    {
      "id": "claude-fable-5-2-max",
      "name": "Claude Fable 5.2",
      "shortLabel": "Fable 5.2 max",
      "effort": "max",
      "vendor": "anthropic",
      "weights": "closed",
      "aaSlug": "claude-fable-5-2-max",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 8,
        "outputPer1M": 40
      },
      "metrics": {
        "aaIndex": 65,
        "aaCostUsd": 26.54,
        "aaReasoningTokensK": 193,
        "aaAnswerTokensK": 81,
        "aaTotalOutputTokensK": 274,
        "aaTimeMin": 41.2,
        "cursorBenchPct": 61.3,
        "cursorCostUsd": 48.11,
        "terminalBenchPct": 65.8,
        "frontierCodePct": null,
        "hallucinationPct": 59.3
      },
      "chartOffset": {
        "aa": {
          "dx": -8,
          "dy": 4,
          "anchor": "end"
        },
        "cursor": {
          "dx": -8,
          "dy": 4,
          "anchor": "end"
        }
      }
    },
    {
      "id": "gemini-4-carbon-high",
      "name": "Gemini 4 Carbon",
      "shortLabel": "Carbon high",
      "effort": "high",
      "vendor": "google",
      "weights": "closed",
      "aaSlug": "gemini-4-carbon-high",
      "showInTokenBreakdown": false,
      "pricing": {
        "inputPer1M": 2,
        "outputPer1M": 10,
        "promo": {
          "regularInputPer1M": 4,
          "regularOutputPer1M": 20,
          "expiresAt": "2026-12-01"
        }
      },
      "metrics": {
        "aaIndex": 60,
        "aaCostUsd": 4.9,
        "aaReasoningTokensK": 96,
        "aaAnswerTokensK": 58,
        "aaTotalOutputTokensK": 154,
        "aaTimeMin": 36.1,
        "cursorBenchPct": 61.5,
        "cursorCostUsd": 8.52,
        "terminalBenchPct": 64.5,
        "frontierCodePct": null,
        "hallucinationPct": 11.4
      },
      "chartOffset": {
        "aa": {
          "dx": 7,
          "dy": 11,
          "anchor": "start"
        },
        "cursor": {
          "dx": 7,
          "dy": -6,
          "anchor": "start"
        }
      }
    },
    {
      "id": "claude-opus-5-5-max",
      "name": "Claude Opus 5.5",
      "shortLabel": "Opus 5.5 max",
      "effort": "max",
      "vendor": "anthropic",
      "weights": "closed",
      "aaSlug": "claude-opus-5-5-max",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 4,
        "outputPer1M": 20
      },
      "metrics": {
        "aaIndex": 59,
        "aaCostUsd": 10.04,
        "aaReasoningTokensK": 120,
        "aaAnswerTokensK": 53,
        "aaTotalOutputTokensK": 173,
        "aaTimeMin": 24.6,
        "cursorBenchPct": 58.1,
        "cursorCostUsd": 11.88,
        "terminalBenchPct": 65.8,
        "frontierCodePct": 47.3,
        "hallucinationPct": 50.3
      },
      "chartOffset": {
        "aa": {
          "dx": 7,
          "dy": 4,
          "anchor": "start"
        },
        "cursor": {
          "dx": 7,
          "dy": 4,
          "anchor": "start"
        }
      }
    },
    {
      "id": "gpt-6-2-sol-max",
      "name": "GPT-6.2 Sol",
      "shortLabel": "Sol max",
      "effort": "max",
      "vendor": "openai",
      "weights": "closed",
      "aaSlug": "gpt-6-2-sol-max",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 2,
        "outputPer1M": 10,
        "contextCapK": 272
      },
      "metrics": {
        "aaIndex": 57,
        "aaCostUsd": 1.01,
        "aaReasoningTokensK": 23,
        "aaAnswerTokensK": 17,
        "aaTotalOutputTokensK": 40,
        "aaTimeMin": 8.5,
        "cursorBenchPct": null,
        "cursorCostUsd": null,
        "terminalBenchPct": 61.5,
        "frontierCodePct": 43.8,
        "hallucinationPct": 55.4
      },
      "chartOffset": {
        "aa": {
          "dx": 7,
          "dy": 4,
          "anchor": "start"
        }
      }
    },
    {
      "id": "claude-sonnet-5-5-max",
      "name": "Claude Sonnet 5.5",
      "shortLabel": "Sonnet 5.5 max",
      "effort": "max",
      "vendor": "anthropic",
      "weights": "closed",
      "aaSlug": "claude-sonnet-5-5-max",
      "showInTokenBreakdown": true,
      "pricing": {
        "inputPer1M": 2,
        "outputPer1M": 10
      },
      "metrics": {
        "aaIndex": 55,
        "aaCostUsd": 10.02,
        "aaReasoningTokensK": 206,
        "aaAnswerTokensK": 79,
        "aaTotalOutputTokensK": 285,
        "aaTimeMin": 22.4,
        "cursorBenchPct": 54.4,
        "cursorCostUsd": 8.63,
        "terminalBenchPct": 63.9,
        "frontierCodePct": null,
        "hallucinationPct": null
      },
      "chartOffset": {
        "aa": {
          "dx": 0,
          "dy": 14,
          "anchor": "middle"
        },
        "cursor": {
          "dx": 0,
          "dy": 14,
          "anchor": "middle"
        }
      }
    },
    {
      "id": "gemini-4-hydrogen-xhigh",
      "name": "Gemini 4 Hydrogen",
      "shortLabel": "Hydrogen xhigh",
      "effort": "xhigh",
      "vendor": "google",
      "weights": "closed",
      "aaSlug": "gemini-4-hydrogen-xhigh",
      "showInTokenBreakdown": false,
      "pricing": {
        "inputPer1M": 1.5,
        "outputPer1M": 7.5
      },
      "metrics": {
        "aaIndex": 54,
        "aaCostUsd": 3.16,
        "aaReasoningTokensK": 68,
        "aaAnswerTokensK": 42,
        "aaTotalOutputTokensK": 110,
        "aaTimeMin": 11.2,
        "cursorBenchPct": 61.4,
        "cursorCostUsd": 8.09,
        "terminalBenchPct": 56.4,
        "terminalBenchNote": "high",
        "frontierCodePct": null,
        "hallucinationPct": 8.5
      },
      "chartOffset": {
        "aa": {
          "dx": 0,
          "dy": 14,
          "anchor": "middle"
        },
        "cursor": {
          "dx": -8,
          "dy": 12,
          "anchor": "end"
        }
      }
    }
  ],
  "picks": {
    "coding": {
      "upstreamSources": [
        "cursorbench",
        "terminal_bench",
        "frontier_code"
      ],
      "cards": [
        {
          "tier": "default",
          "modelId": "gemini-4-neon-high",
          "primaryMetricKey": "cursorBenchPct",
          "primaryMetricSourceId": "cursorbench",
          "costMetricKey": "cursorCostUsd"
        },
        {
          "tier": "budget",
          "modelId": "gpt-6-2-sol-max",
          "primaryMetricKey": "terminalBenchPct",
          "primaryMetricSourceId": "terminal_bench",
          "costMetricKey": "aaCostUsd"
        },
        {
          "tier": "max",
          "modelId": "gemini-4-neon-xhigh",
          "primaryMetricKey": "cursorBenchPct",
          "primaryMetricSourceId": "cursorbench",
          "costMetricKey": "cursorCostUsd"
        }
      ]
    },
    "office": {
      "upstreamSources": [
        "aa_index",
        "deepmind_eval",
        "aa_omniscience"
      ],
      "cards": [
        {
          "tier": "default",
          "modelId": "gemini-4-neon-high",
          "primaryMetricKey": "automationBenchPct",
          "primaryMetricSourceId": "deepmind_eval",
          "costMetricKey": "aaCostUsd"
        },
        {
          "tier": "budget",
          "modelId": "gemini-4-hydrogen-xhigh",
          "primaryMetricKey": "hallucinationPct",
          "primaryMetricSourceId": "aa_omniscience",
          "costMetricKey": "aaCostUsd"
        },
        {
          "tier": "max",
          "modelId": "gpt-6-2-astra-max",
          "primaryMetricKey": "aaIndex",
          "primaryMetricSourceId": "aa_index",
          "costMetricKey": "aaCostUsd"
        }
      ]
    },
    "general": {
      "upstreamSources": [
        "aa_index",
        "aa_omniscience",
        "aa_models"
      ],
      "cards": [
        {
          "tier": "default",
          "modelId": "gemini-4-neon-high",
          "primaryMetricKey": "aaIndex",
          "primaryMetricSourceId": "aa_index",
          "costMetricKey": "aaCostUsd"
        },
        {
          "tier": "budget",
          "modelId": "gpt-6-2-sol-max",
          "primaryMetricKey": "aaIndex",
          "primaryMetricSourceId": "aa_index",
          "costMetricKey": "aaCostUsd"
        },
        {
          "tier": "max",
          "modelId": "claude-fable-5-2-max",
          "primaryMetricKey": "aaIndex",
          "primaryMetricSourceId": "aa_index",
          "costMetricKey": "aaCostUsd"
        }
      ]
    }
  },
  "editorial": {
    "en": {
      "selectorLabel": "2026-11-18 · Neon Release Reset",
      "headline": "Gemini 4 Neon Resets Coding Cost Frontier",
      "seoDescription": "2026-11-18 LLM real cost snapshot comparing Gemini 4 Neon, GPT-6.2 Astra/Sol, and Claude 5.5/Fable across per-task USD spend, reasoning tokens, and benchmarks.",
      "brief": [
        {
          "code": "01",
          "text": "Gemini 4 Neon takes the top two spots on CursorBench (65.4% / 63.7%). Its high tier averages $6.54 per task, 45% below Opus 5.5."
        },
        {
          "code": "02",
          "text": "GPT-6.2 Astra and Sol compress per-task output to ~40k tokens. Despite an $8/$40 list price, Astra averages $4.00 per AA task."
        },
        {
          "code": "03",
          "text": "Gemini 4 Carbon introductory half-price API ($2/$10) ends Nov 30, reverting to $4/$20 on Dec 1."
        }
      ],
      "scatterNotes": {
        "cursor": "GPT-6.2 series is not yet listed on CursorBench 4.1."
      },
      "cards": {
        "coding.default": {
          "delta": "NEW default",
          "basis": "Ranks 2nd on CursorBench (63.7%) at 55% of Opus 5.5 task spend.",
          "boundary": "Trails Opus by 0.6 pts on pure terminal tasks; peak SWE relies on official CLI."
        },
        "coding.budget": {
          "delta": "NEW lowest bill",
          "basis": "40k output budget produces the lowest per-task spend for routine scripts.",
          "boundary": "Input above 272K triggers higher pricing; 55.4% open-fact hallucination rate."
        },
        "coding.max": {
          "delta": "NEW #1 (65.4%)",
          "basis": "Highest multi-file refactor score with 65.2% on Terminal-Bench.",
          "boundary": "Costs 63% more per task than high tier for a 1.7 pt score gain."
        },
        "office.default": {
          "delta": "-63% cost vs Opus",
          "basis": "Leads cross-app office automation at 56.3% with 13.9% hallucination rate.",
          "boundary": "Trails GPT-6.2 Astra on complex financial spreadsheet modeling."
        },
        "office.budget": {
          "delta": "Lowest hallucination",
          "basis": "Strong long-PDF parsing with the lowest hallucination rate in the set (8.5%).",
          "boundary": "Multi-step reasoning depth trails flagship configurations."
        },
        "office.max": {
          "delta": "NEW modeling lead",
          "basis": "Leads financial modeling and dossier synthesis at 13.7m average task time.",
          "boundary": "Low abstention on unknown facts results in a 59.1% hallucination rate."
        },
        "general.default": {
          "delta": "Low-hallucination pick",
          "basis": "Reliable 1M-context retrieval with strong factual calibration.",
          "boundary": "Carbon $2/$10 promotional pricing expires after Nov 30."
        },
        "general.budget": {
          "delta": "NEW fastest bill",
          "basis": "Finishes tasks in 8.5m average with the lowest recurring Q&A bill.",
          "boundary": "Prone to guessing on niche facts; pair with web search for verification."
        },
        "general.max": {
          "delta": "AA #1 (65)",
          "basis": "Highest AA Intelligence Index (65) for deep analytical reasoning.",
          "boundary": "Averages $26.54 and 41.2m per task due to 274k output tokens."
        }
      }
    },
    "zh": {
      "selectorLabel": "2026-11-18 · Neon 发布洗牌",
      "headline": "Gemini 4 Neon 重塑编码真实成本前沿",
      "seoDescription": "2026-11-18 大模型选型与真实成本快照：对比 Gemini 4 Neon、GPT-6.2、Claude Fable/Opus 的思考 Token 消耗与多榜实测成绩。",
      "brief": [
        {
          "code": "01",
          "text": "Gemini 4 Neon 包揽 CursorBench 冠亚军（65.4% / 63.7%），high 档单次均价 $6.54，较 Opus 5.5 降低 45%。"
        },
        {
          "code": "02",
          "text": "GPT-6.2 Astra 与 Sol 单次输出压缩至约 40k tokens，Astra 虽标价 $8/$40，AA 单次实际均价仅 $4.00。"
        },
        {
          "code": "03",
          "text": "Gemini 4 Carbon 首发半价 API（$2/$10）将于 11 月 30 日截止，12 月 1 日起恢复 $4/$20。"
        }
      ],
      "scatterNotes": {
        "cursor": "CursorBench 4.1 暂未收录 GPT-6.2 系列。"
      },
      "cards": {
        "coding.default": {
          "delta": "NEW 默认主力",
          "basis": "CursorBench 63.7% 排第二，账单仅 Opus 55%。",
          "boundary": "纯终端较 Opus 低 0.6%；SWE 高分依赖官方 CLI。"
        },
        "coding.budget": {
          "delta": "NEW 单次最低",
          "basis": "输出精炼至 40k，常规脚本单次账单最低。",
          "boundary": "输入超 272K 单价跳涨；开放事实幻觉率 55.4%。"
        },
        "coding.max": {
          "delta": "NEW 榜首 65.4%",
          "basis": "跨文件重构第一，Terminal-Bench 达 65.2%。",
          "boundary": "较 high 档成本增加 63%，得分提升 1.7 个百分点。"
        },
        "office.default": {
          "delta": "成本 -63% vs Opus",
          "basis": "跨软件办公 56.3% 排第一，幻觉率仅 13.9%。",
          "boundary": "复杂财务表格建模得分不及 Astra。"
        },
        "office.budget": {
          "delta": "长文档稳健",
          "basis": "长文档解析居前，幻觉率全场最低（8.5%）。",
          "boundary": "多步复杂逻辑推演弱于旗舰梯队。"
        },
        "office.max": {
          "delta": "NEW 建模第一",
          "basis": "财务建模与案卷规划领先，单次仅耗时 13.7m。",
          "boundary": "未知事实拒答率偏低，幻觉率达 59.1%。"
        },
        "general.default": {
          "delta": "低幻觉主力",
          "basis": "1M 上下文全量检索稳健，事实准确率高。",
          "boundary": "Carbon $2/$10 优惠价 11 月 30 日截止。"
        },
        "general.budget": {
          "delta": "NEW 极速响应",
          "basis": "单次平均 8.5m 完结，高频问答账单最低。",
          "boundary": "冷门知识易硬猜，严肃核查需配搜索。"
        },
        "general.max": {
          "delta": "综合分 65 第一",
          "basis": "AA 综合指数 65 分第一，适合深度推演。",
          "boundary": "单次均价 $26.54 且平均耗时达 41.2m。"
        }
      }
    },
    "ja": {
      "selectorLabel": "2026-11-18 · Neon 公開後の再編",
      "headline": "Gemini 4 Neon が開発実費フロンティアを更新",
      "seoDescription": "2026-11-18 時点の主要 LLM 実質コスト比較。思考トークンを含む1回実費と主要ベンチマークを Gemini 4 Neon、GPT-6.2、Claude で横断比較。",
      "brief": [
        {
          "code": "01",
          "text": "Gemini 4 Neon が CursorBench 1・2位（65.4% / 63.7%）を獲得。high 設定の1回実費は $6.54 で Opus 5.5 比 45% 低減。"
        },
        {
          "code": "02",
          "text": "GPT-6.2 Astra と Sol は1回出力を約 40k トークンに抑制。Astra は定価 $8/$40 ながら AA 実費は $4.00。"
        },
        {
          "code": "03",
          "text": "Gemini 4 Carbon の初回半額 API（$2/$10）は11月30日終了、12月1日より $4/$20 に改定。"
        }
      ],
      "scatterNotes": {
        "cursor": "CursorBench 4.1 には GPT-6.2 シリーズが未掲載。"
      },
      "cards": {
        "coding.default": {
          "delta": "NEW 標準",
          "basis": "CursorBench 63.7% で2位、Opus 5.5 の 55% の実費。",
          "boundary": "純ターミナルは Opus に 0.6pt 劣後。SWE 上限は公式 CLI 依存。"
        },
        "coding.budget": {
          "delta": "NEW 最小実費",
          "basis": "出力 40k に収まり定常スクリプトの1回あたり実費が最小。",
          "boundary": "272K 入力超で単価上昇。一般事実の誤答率 55.4%。"
        },
        "coding.max": {
          "delta": "NEW 首位 65.4%",
          "basis": "複数ファイル改修で最高点、Terminal-Bench も 65.2%。",
          "boundary": "high 設定比で実費 +63% に対しスコア上昇は 1.7pt。"
        },
        "office.default": {
          "delta": "Opus 比 -63%",
          "basis": "複数アプリ連携業務 56.3% で首位、誤答率 13.9%。",
          "boundary": "複雑な財務モデリングでは GPT-6.2 Astra に劣後。"
        },
        "office.budget": {
          "delta": "最小誤答率",
          "basis": "長文 PDF 解析で安定し、誤答率は全体最小の 8.5%。",
          "boundary": "多段階の複雑推論は最上位モデル群より控えめ。"
        },
        "office.max": {
          "delta": "NEW 財務推論首位",
          "basis": "財務モデリングと資料設計で先行し、平均所要時間は 13.7m。",
          "boundary": "未知事実の保留率が低く、誤答率は 59.1%。"
        },
        "general.default": {
          "delta": "低誤答の定番",
          "basis": "1M コンテキスト検索が安定し、事実精度との均衡が高い。",
          "boundary": "Carbon の $2/$10 割引価格は11月30日で終了。"
        },
        "general.budget": {
          "delta": "NEW 高速・低コスト",
          "basis": "平均 8.5m で完了し、高頻度 Q&A の実費が最小。",
          "boundary": "専門外の事実で推測が出やすく、厳密な検証には検索併用が必要。"
        },
        "general.max": {
          "delta": "AA 総合 65 首位",
          "basis": "AA 総合指数 65 で首位、深い多段推論向け。",
          "boundary": "274k 出力により1回平均 $26.54・所要 41.2m。"
        }
      }
    }
  },
  "computed": {
    "paretoAA": [
      {
        "id": "gpt-6-2-sol-max",
        "name": "GPT-6.2 Sol (max)",
        "aaIndex": 57,
        "aaCostUsd": 1.01
      },
      {
        "id": "gemini-4-neon-high",
        "name": "Gemini 4 Neon (high)",
        "aaIndex": 60,
        "aaCostUsd": 3.74
      },
      {
        "id": "gpt-6-2-astra-max",
        "name": "GPT-6.2 Astra (max)",
        "aaIndex": 62,
        "aaCostUsd": 4
      },
      {
        "id": "claude-fable-5-2-max",
        "name": "Claude Fable 5.2 (max)",
        "aaIndex": 65,
        "aaCostUsd": 26.54
      }
    ],
    "paretoCursor": [
      {
        "id": "gemini-4-neon-high",
        "name": "Gemini 4 Neon (high)",
        "cursorBenchPct": 63.7,
        "cursorCostUsd": 6.54
      },
      {
        "id": "gemini-4-neon-xhigh",
        "name": "Gemini 4 Neon (xhigh)",
        "cursorBenchPct": 65.4,
        "cursorCostUsd": 10.67
      }
    ],
    "paretoTerminal": [
      {
        "id": "gpt-6-2-sol-max",
        "name": "GPT-6.2 Sol (max)",
        "terminalBenchPct": 61.5,
        "aaCostUsd": 1.01
      },
      {
        "id": "gpt-6-2-astra-max",
        "name": "GPT-6.2 Astra (max)",
        "terminalBenchPct": 62.1,
        "aaCostUsd": 4
      },
      {
        "id": "gemini-4-carbon-high",
        "name": "Gemini 4 Carbon (high)",
        "terminalBenchPct": 64.5,
        "aaCostUsd": 4.9
      },
      {
        "id": "gemini-4-neon-xhigh",
        "name": "Gemini 4 Neon (xhigh)",
        "terminalBenchPct": 65.2,
        "aaCostUsd": 6.44
      },
      {
        "id": "claude-opus-5-5-max",
        "name": "Claude Opus 5.5 (max)",
        "terminalBenchPct": 65.8,
        "aaCostUsd": 10.04
      }
    ]
  },
  "sources": {
    "aa_index": {
      "id": "aa_index",
      "name": "Artificial Analysis Intelligence Index v5.0.1",
      "shortName": "Artificial Analysis v5.0.1",
      "org": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index",
      "version": "v5.0.1",
      "accessMethod": "api_free_plus_public"
    },
    "aa_models": {
      "id": "aa_models",
      "name": "Artificial Analysis Model Leaderboard",
      "shortName": "Artificial Analysis",
      "org": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/leaderboards/models/",
      "version": "v5.0.1",
      "accessMethod": "api_free"
    },
    "aa_omniscience": {
      "id": "aa_omniscience",
      "name": "AA-Omniscience v1.1 Hallucination Benchmark",
      "shortName": "AA-Omniscience v1.1",
      "org": "Artificial Analysis",
      "url": "https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index",
      "version": "v1.1",
      "accessMethod": "public_web"
    },
    "cursorbench": {
      "id": "cursorbench",
      "name": "CursorBench 4.1",
      "shortName": "CursorBench 4.1",
      "org": "Anysphere",
      "url": "https://cursor.com/cursorbench",
      "version": "4.1",
      "accessMethod": "public_web"
    },
    "terminal_bench": {
      "id": "terminal_bench",
      "name": "Terminal-Bench 4.1",
      "shortName": "Terminal-Bench 4.1",
      "org": "Snorkel AI / Harbor",
      "url": "https://snorkel.ai/leaderboard/terminal-bench/",
      "version": "4.1",
      "accessMethod": "public_web"
    },
    "frontier_code": {
      "id": "frontier_code",
      "name": "FrontierCode 2.0",
      "shortName": "FrontierCode 2.0",
      "org": "Cognition",
      "url": "https://cognition.com/blog/frontier-code-2.0",
      "version": "2.0",
      "accessMethod": "public_web"
    },
    "deepmind_eval": {
      "id": "deepmind_eval",
      "name": "Google DeepMind Model Evaluation Report",
      "shortName": "DeepMind Eval",
      "org": "Google DeepMind",
      "url": "https://deepmind.google/models/evals-methodology/gemini-4-neon",
      "version": "2026-11",
      "accessMethod": "vendor_report"
    }
  }
}