{
  "metadata": {
    "project": "SXF / AI",
    "reviewed_at": "2026-11-13",
    "kind": "source-verifiable benchmark methodology transition",
    "warning": "The comparison is about methodology versions, not time-series model performance; no model scores inferred.",
    "aa_v43_source": "https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v4-3",
    "aa_v5_source": "https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5/"
  },
  "versions": [
    {
      "version": "v4.3",
      "released": "2026-09-07",
      "evaluations": 10,
      "category_weights_pct": {
        "Agents": 30,
        "Coding": 20,
        "General": 30,
        "Scientific reasoning": 20
      },
      "private_evaluation_weight_pct": 45,
      "source": "https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v4-3"
    },
    {
      "version": "v5.0.1",
      "released": "2026-10-27 (v5 introduction); v5.0.1 as reviewed 2026-11-13",
      "evaluations": 9,
      "category_weights_pct": {
        "Knowledge Work": 25,
        "Coding": 25,
        "Science": 25,
        "General": 25
      },
      "private_evaluation_weight_pct": 60,
      "source": "https://artificialanalysis.ai/methodology/intelligence-benchmarking"
    }
  ],
  "v5_changes": {
    "new": [
      "FrontierCode 2.0",
      "Terminal-Bench-Science 0.1.0"
    ],
    "retired_from_index": [
      "GDPval-AA v2.1",
      "SciCode",
      "AA-LCR v1.1"
    ],
    "upgraded": [
      "HLE-Diamond",
      "AA-Omniscience v1.1",
      "CritPt v2 hard subset"
    ],
    "retained": [
      "AA-Briefcase v1.1",
      "AutomationBench-AA",
      "GDP.pdf",
      "Terminal-Bench 4.0"
    ],
    "source": "https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5/"
  }
}