{
  "name": "Aeon Model Economics Lab dataset",
  "version": "2026-07-10",
  "updated": "2026-07-10",
  "license": "Aeon research compilation. Source benchmark terms continue to apply.",
  "methodology": "Coding Agent Index on the y-axis, task cost on the x-axis, and output tokens encoded as bubble area. Measured and modeled points are explicitly separated.",
  "sources": [
    {
      "label": "OpenAI GPT-5.6 launch and pricing",
      "url": "https://openai.com/index/gpt-5-6/"
    },
    {
      "label": "Artificial Analysis Coding Agent Index",
      "url": "https://artificialanalysis.ai/agents/coding-agents"
    },
    {
      "label": "Artificial Analysis GPT-5.6 analysis",
      "url": "https://artificialanalysis.ai/articles/gpt-5-6-has-landed"
    },
    {
      "label": "OpenAI GPT-5.5 launch",
      "url": "https://openai.com/index/introducing-gpt-5-5/"
    },
    {
      "label": "xAI Grok 4.5 announcement",
      "url": "https://x.ai/news/grok-4-5"
    },
    {
      "label": "Anthropic Claude Fable 5",
      "url": "https://www.anthropic.com/claude/fable"
    },
    {
      "label": "Anthropic effort documentation",
      "url": "https://platform.claude.com/docs/en/build-with-claude/effort"
    },
    {
      "label": "Agentic coding token study",
      "url": "https://arxiv.org/abs/2604.22750"
    }
  ],
  "series": [
    {
      "id": "sol",
      "name": "GPT-5.6 Sol",
      "shortName": "Sol",
      "provider": "OpenAI",
      "color": "#E3B341",
      "points": [
        {
          "effort": "Low",
          "score": 43,
          "outputTokens": 3500,
          "cost": 0.65,
          "evidence": "modeled",
          "basis": "Estimated from neighboring measured GPT-5.6 family effort curves.",
          "sourceUrl": "https://artificialanalysis.ai/articles/gpt-5-6-has-landed"
        },
        {
          "effort": "Medium",
          "score": 52,
          "outputTokens": 5500,
          "cost": 0.95,
          "evidence": "modeled",
          "basis": "Estimated from neighboring measured GPT-5.6 family effort curves.",
          "sourceUrl": "https://artificialanalysis.ai/articles/gpt-5-6-has-landed"
        },
        {
          "effort": "High",
          "score": 58.4,
          "outputTokens": 7663,
          "cost": 1.398,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "XHigh",
          "score": 71.8,
          "outputTokens": 31662,
          "cost": 1.59,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Max",
          "score": 80,
          "outputTokens": 54860,
          "cost": 7.084,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    },
    {
      "id": "terra",
      "name": "GPT-5.6 Terra",
      "shortName": "Terra",
      "provider": "OpenAI",
      "color": "#4FB6C2",
      "points": [
        {
          "effort": "Low",
          "score": 53.8,
          "outputTokens": 8053,
          "cost": 0.484,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Medium",
          "score": 64.2,
          "outputTokens": 16002,
          "cost": 0.901,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "High",
          "score": 69.2,
          "outputTokens": 25400,
          "cost": 1.31,
          "evidence": "modeled",
          "basis": "Monotone interpolation between measured Terra medium and xhigh runs.",
          "sourceUrl": "https://artificialanalysis.ai/articles/gpt-5-6-has-landed"
        },
        {
          "effort": "XHigh",
          "score": 73.2,
          "outputTokens": 40160,
          "cost": 1.895,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Max",
          "score": 77.4,
          "outputTokens": 60732,
          "cost": 2.756,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    },
    {
      "id": "luna",
      "name": "GPT-5.6 Luna",
      "shortName": "Luna",
      "provider": "OpenAI",
      "color": "#42B883",
      "points": [
        {
          "effort": "Low",
          "score": 42.4,
          "outputTokens": 6699,
          "cost": 0.209,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Medium",
          "score": 58.7,
          "outputTokens": 15198,
          "cost": 0.474,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "High",
          "score": 67.9,
          "outputTokens": 32273,
          "cost": 0.959,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "XHigh",
          "score": 70.8,
          "outputTokens": 47114,
          "cost": 1.256,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Max",
          "score": 74.6,
          "outputTokens": 64941,
          "cost": 1.567,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    },
    {
      "id": "gpt55",
      "name": "GPT-5.5",
      "shortName": "GPT-5.5",
      "provider": "OpenAI",
      "color": "#CBD5E1",
      "points": [
        {
          "effort": "Low",
          "score": 54.5,
          "outputTokens": 8100,
          "cost": 1.45,
          "evidence": "modeled",
          "basis": "Estimated from the measured GPT-5.5 medium and xhigh runs and GPT-5.6 effort curves.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "Medium",
          "score": 70.5,
          "outputTokens": 17196,
          "cost": 2.755,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "High",
          "score": 73.5,
          "outputTokens": 26000,
          "cost": 3.75,
          "evidence": "modeled",
          "basis": "Monotone interpolation between measured GPT-5.5 medium and xhigh runs.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        },
        {
          "effort": "XHigh",
          "score": 76.4,
          "outputTokens": 37947,
          "cost": 5.07,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    },
    {
      "id": "grok",
      "name": "Grok 4.5",
      "shortName": "Grok",
      "provider": "xAI",
      "color": "#EF7C5A",
      "points": [
        {
          "effort": "Low",
          "score": 56.2,
          "outputTokens": 12700,
          "cost": 0.78,
          "evidence": "modeled",
          "basis": "Estimated by applying the median observed GPT-5.6 low-to-high effort shape to the measured Grok high anchor.",
          "sourceUrl": "https://x.ai/news/grok-4-5"
        },
        {
          "effort": "Medium",
          "score": 68,
          "outputTokens": 25200,
          "cost": 1.56,
          "evidence": "modeled",
          "basis": "Estimated by applying the median observed GPT-5.6 medium-to-high effort shape to the measured Grok high anchor.",
          "sourceUrl": "https://x.ai/news/grok-4-5"
        },
        {
          "effort": "High",
          "score": 76.4,
          "outputTokens": 39966,
          "cost": 2.594,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    },
    {
      "id": "fable",
      "name": "Claude Fable 5",
      "shortName": "Fable",
      "provider": "Anthropic",
      "color": "#B99AF8",
      "points": [
        {
          "effort": "Low",
          "score": 28.5,
          "outputTokens": 33340,
          "cost": 3.27,
          "evidence": "modeled",
          "basis": "Anthropic FrontierCode effort-curve shape normalized to the measured Fable max Coding Agent Index anchor.",
          "sourceUrl": "https://platform.claude.com/docs/en/build-with-claude/effort"
        },
        {
          "effort": "Medium",
          "score": 44.5,
          "outputTokens": 43920,
          "cost": 4.32,
          "evidence": "modeled",
          "basis": "Anthropic FrontierCode effort-curve shape normalized to the measured Fable max Coding Agent Index anchor.",
          "sourceUrl": "https://platform.claude.com/docs/en/build-with-claude/effort"
        },
        {
          "effort": "High",
          "score": 60,
          "outputTokens": 61350,
          "cost": 6.03,
          "evidence": "modeled",
          "basis": "Anthropic FrontierCode effort-curve shape normalized to the measured Fable max Coding Agent Index anchor.",
          "sourceUrl": "https://platform.claude.com/docs/en/build-with-claude/effort"
        },
        {
          "effort": "XHigh",
          "score": 73,
          "outputTokens": 76620,
          "cost": 7.53,
          "evidence": "modeled",
          "basis": "Anthropic FrontierCode effort-curve shape normalized to the measured Fable max Coding Agent Index anchor.",
          "sourceUrl": "https://platform.claude.com/docs/en/build-with-claude/effort"
        },
        {
          "effort": "Max",
          "score": 77.2,
          "outputTokens": 119611,
          "cost": 11.75,
          "evidence": "measured",
          "basis": "Measured Artificial Analysis Coding Agent Index run with fallback configuration, accessed July 10, 2026.",
          "sourceUrl": "https://artificialanalysis.ai/agents/coding-agents"
        }
      ]
    }
  ]
}