{
  "schema": "builderwars.eval-source-extract.v1",
  "id": "matharena-expected",
  "sourceUrl": "https://matharena.ai/",
  "sourceSha256": "6815e5a1bd93bb2780937c309bea1ba284d26eedd42d0abf4eb49cc64403e0c1",
  "capturedAt": "2026-10-07T07:01:28Z",
  "extraction": "Retained reported score columns; not complete source page or execution evidence",
  "rows": [
    {
      "name": "GPT-6.1 Sol (max)",
      "score": 90.1,
      "detail": "#1 overall · OpenAI · source uncertainty ±2.3 percentage points · expected cost $0.55 ±$0.045",
      "date": "2026-09-29",
      "rank": 1,
      "sourceRowId": "1"
    },
    {
      "name": "GPT-6 Astra (max)",
      "score": 88.0,
      "detail": "#2 overall · OpenAI · source uncertainty ±2.0 percentage points · expected cost $2.42 ±$0.28",
      "date": "2026-09-04",
      "rank": 2,
      "sourceRowId": "2"
    },
    {
      "name": "GPT-6 Sol (max)",
      "score": 84.9,
      "detail": "#3 overall · OpenAI · source uncertainty ±2.5 percentage points · expected cost $1.57 ±$0.17",
      "date": "2026-09-22",
      "rank": 3,
      "sourceRowId": "3"
    },
    {
      "name": "Qwen3.8-Max",
      "score": 56.1,
      "detail": "Best open model · Qwen · source uncertainty ±7.6 percentage points · expected cost $3.07 ±$0.35",
      "date": "2026-08-02",
      "rank": null,
      "sourceRowId": "4"
    }
  ]
}