{
 "metadata": {
  "name": "AI effective cost per answer 2026",
  "description": "What a finished answer costs, not what a token costs: 9 Claude models answering two batteries (60 ordinary requests and 120 requests across code, summarising, long-form writing and reasoning), priced at the provider-stated list prices archived by AI Price Watch, with the change in output tokens split into tokenizer, visible writing and hidden thinking.",
  "edition": "2026-10-28",
  "doi": "10.5281/zenodo.23748727",
  "conceptDoi": "10.5281/zenodo.23748726",
  "license": "CC-BY-4.0",
  "author": "Vincent Wesley Couey",
  "priceSource": {
   "archive": "lattice-core/_evidence/ai-price-watch-2026/2026-10-26",
   "generated": "2026-10-26T18:47:04.462Z"
  },
  "limits": [
   "Battery 2 is 120 constructed requests (30 each of code, summarising, long-form explanation and short reasoning), not real user prompts: task templates crossed with topic lists, with real article paragraphs as the passages to summarise. It was added because the first battery of 60 was thin and its result did not hold in size on a different mix.",
   "Cost per answer is the measured mean input and output tokens priced at list price. It is not a bill: caching, batch discounts, reasoning tokens and plan limits are ignored.",
   "It measures what answers cost, not how good they are. A longer answer can be a better answer; this dataset does not say.",
   "One battery of 60 ordinary requests, one capture per model per run, Claude models only. Run 1 is the frozen Decay Index capture (1,200-token cap, so some long answers were cut off); run 2 raised the cap to 4,000 and is the headline.",
   "Run 2 was captured before this method note was written, so the method was not committed in advance and this is not a pre-registered study.",
   "Temperature was 0 where the model accepted it and the provider default otherwise; it is recorded per answer. There is no clean run-to-run noise estimate because every model was cut off at least once in run 1; uncertainty is the paired bootstrap interval over prompts.",
   "Hidden thinking tokens are billed inside output_tokens. Their share was measured on a separate recapture of the 5.x models only; for the three oldest models the API reports no thinking count and thinking is taken as 0 because no thinking block was returned. The cost figures use the provider-reported output tokens, so they include thinking.",
   "Claude Haiku 5.5 has no list price in the archived tables, so it has token counts but no cost per answer; no price was guessed.",
   "Only input and output tokens of this battery are priced. The ratio between input and output depends on the task; short prompts with long answers (as here) make output price dominate."
  ]
 },
 "designs": [
  {
   "id": "b1-cap1200",
   "battery": "60 ordinary requests",
   "cap": 1200,
   "run": "frozen Decay Index capture"
  },
  {
   "id": "b1-cap4000",
   "battery": "60 ordinary requests",
   "cap": 4000,
   "run": "run 2 (cap raised)"
  },
  {
   "id": "b2-cap4000",
   "battery": "120 task-typed requests",
   "cap": 4000,
   "run": "battery 2"
  }
 ],
 "headlineRanges": [
  {
   "from": "claude-opus-4-6",
   "to": "claude-opus-4-7",
   "label": "Claude Opus 4.6 to Claude Opus 4.7",
   "priced": true,
   "byDesign": [
    {
     "design": "b1-cap1200",
     "costPerAnswerChangePct": 43.36,
     "cutOffAnswers": 9,
     "worstModelCutOffShare": 0.083
    },
    {
     "design": "b1-cap4000",
     "costPerAnswerChangePct": 26.018,
     "cutOffAnswers": 1,
     "worstModelCutOffShare": 0.017
    },
    {
     "design": "b2-cap4000",
     "costPerAnswerChangePct": 38.83,
     "cutOffAnswers": 0,
     "worstModelCutOffShare": 0
    }
   ],
   "headlineDesigns": [
    "b1-cap4000",
    "b2-cap4000"
   ],
   "minPct": 26.018,
   "maxPct": 38.83,
   "spreadPts": 12.812,
   "signFlipsAcrossDesigns": false,
   "varied": "the battery (60 ordinary requests against 120 task-typed requests)"
  },
  {
   "from": "claude-opus-4-7",
   "to": "claude-opus-4-8",
   "label": "Claude Opus 4.7 to Claude Opus 4.8",
   "priced": true,
   "byDesign": [
    {
     "design": "b1-cap1200",
     "costPerAnswerChangePct": -0.608,
     "cutOffAnswers": 8,
     "worstModelCutOffShare": 0.083
    },
    {
     "design": "b1-cap4000",
     "costPerAnswerChangePct": -1.637,
     "cutOffAnswers": 0,
     "worstModelCutOffShare": 0
    },
    {
     "design": "b2-cap4000",
     "costPerAnswerChangePct": -0.714,
     "cutOffAnswers": 0,
     "worstModelCutOffShare": 0
    }
   ],
   "headlineDesigns": [
    "b1-cap4000",
    "b2-cap4000"
   ],
   "minPct": -1.637,
   "maxPct": -0.714,
   "spreadPts": 0.923,
   "signFlipsAcrossDesigns": false,
   "varied": "the battery (60 ordinary requests against 120 task-typed requests)"
  },
  {
   "from": "claude-opus-4-6",
   "to": "claude-opus-5-5",
   "label": "Claude Opus 4.6 to Claude Opus 5.5",
   "priced": true,
   "byDesign": [
    {
     "design": "b1-cap1200",
     "costPerAnswerChangePct": 72.525,
     "cutOffAnswers": 24,
     "worstModelCutOffShare": 0.333
    },
    {
     "design": "b1-cap4000",
     "costPerAnswerChangePct": 90.877,
     "cutOffAnswers": 3,
     "worstModelCutOffShare": 0.033
    },
    {
     "design": "b2-cap4000",
     "costPerAnswerChangePct": 37.596,
     "cutOffAnswers": 0,
     "worstModelCutOffShare": 0
    }
   ],
   "headlineDesigns": [
    "b1-cap4000",
    "b2-cap4000"
   ],
   "minPct": 37.596,
   "maxPct": 90.877,
   "spreadPts": 53.281,
   "signFlipsAcrossDesigns": false,
   "varied": "the battery (60 ordinary requests against 120 task-typed requests)"
  },
  {
   "from": "claude-sonnet-4-6",
   "to": "claude-sonnet-5-5",
   "label": "Claude Sonnet 4.6 to Claude Sonnet 5.5",
   "priced": true,
   "byDesign": [
    {
     "design": "b1-cap1200",
     "costPerAnswerChangePct": 7.146,
     "cutOffAnswers": 15,
     "worstModelCutOffShare": 0.15
    },
    {
     "design": "b1-cap4000",
     "costPerAnswerChangePct": 4.735,
     "cutOffAnswers": 2,
     "worstModelCutOffShare": 0.017
    },
    {
     "design": "b2-cap4000",
     "costPerAnswerChangePct": -15.508,
     "cutOffAnswers": 0,
     "worstModelCutOffShare": 0
    }
   ],
   "headlineDesigns": [
    "b1-cap4000",
    "b2-cap4000"
   ],
   "minPct": -15.508,
   "maxPct": 4.735,
   "spreadPts": 20.243,
   "signFlipsAcrossDesigns": true,
   "varied": "the battery (60 ordinary requests against 120 task-typed requests)"
  },
  {
   "from": "claude-haiku-4-5-20251001",
   "to": "claude-haiku-5-5",
   "priced": false
  }
 ],
 "closure": {
  "note": "Checks that the parts rebuild the whole. A residual is published, not hidden.",
  "answersCsvRows": {
   "expected": 2160,
   "actual": 2160
  },
  "threeWaySplitVsProvider": [
   {
    "step": "claude-opus-4-6 to claude-opus-4-7",
    "threeWayCombinedPct": 25.706,
    "providerReportedPct": 25.76,
    "residualPts": -0.054
   },
   {
    "step": "claude-opus-4-7 to claude-opus-4-8",
    "threeWayCombinedPct": -1.13,
    "providerReportedPct": -1.49,
    "residualPts": 0.36
   },
   {
    "step": "claude-opus-4-6 to claude-opus-5-5",
    "threeWayCombinedPct": 135.936,
    "providerReportedPct": 139.88,
    "residualPts": -3.944
   },
   {
    "step": "claude-sonnet-4-6 to claude-sonnet-5-5",
    "threeWayCombinedPct": 60.823,
    "providerReportedPct": 57.34,
    "residualPts": 3.483
   },
   {
    "step": "claude-haiku-4-5-20251001 to claude-haiku-5-5",
    "threeWayCombinedPct": 137.457,
    "providerReportedPct": 137.46,
    "residualPts": -0.003
   }
  ],
  "thinkingNeverExceedsBilledOutput": true
 },
 "models": [
  {
   "id": "claude-haiku-4-5-20251001",
   "display": "Claude Haiku 4.5",
   "family": "haiku",
   "listPriceUsdPerMTok": {
    "input": 1,
    "output": 5
   },
   "run1": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 286.883,
    "truncated": 2,
    "costPerAnswerUsd": 0.001462
   },
   "run2": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 317.5,
    "truncated": 0,
    "costPerAnswerUsd": 0.001615
   }
  },
  {
   "id": "claude-haiku-5-5",
   "display": "Claude Haiku 5.5",
   "family": "haiku",
   "listPriceUsdPerMTok": null,
   "run1": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 569.95,
    "truncated": 12,
    "costPerAnswerUsd": null
   },
   "run2": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 753.95,
    "truncated": 1,
    "costPerAnswerUsd": null
   }
  },
  {
   "id": "claude-sonnet-4-6",
   "display": "Claude Sonnet 4.6",
   "family": "sonnet",
   "listPriceUsdPerMTok": {
    "input": 3,
    "output": 15
   },
   "run1": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 391.85,
    "truncated": 6,
    "costPerAnswerUsd": 0.005961
   },
   "run2": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 490.067,
    "truncated": 1,
    "costPerAnswerUsd": 0.007434
   }
  },
  {
   "id": "claude-sonnet-5-5",
   "display": "Claude Sonnet 5.5",
   "family": "sonnet",
   "listPriceUsdPerMTok": {
    "input": 2,
    "output": 10
   },
   "run1": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 631.167,
    "truncated": 9,
    "costPerAnswerUsd": 0.006387
   },
   "run2": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 771.05,
    "truncated": 1,
    "costPerAnswerUsd": 0.007786
   }
  },
  {
   "id": "claude-opus-4-6",
   "display": "Claude Opus 4.6",
   "family": "opus",
   "listPriceUsdPerMTok": {
    "input": 5,
    "output": 25
   },
   "run1": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 366.15,
    "truncated": 4,
    "costPerAnswerUsd": 0.009292
   },
   "run2": {
    "n": 60,
    "meanInTok": 27.65,
    "meanOutTok": 444.3,
    "truncated": 1,
    "costPerAnswerUsd": 0.011246
   }
  },
  {
   "id": "claude-opus-4-7",
   "display": "Claude Opus 4.7",
   "family": "opus",
   "listPriceUsdPerMTok": {
    "input": 5,
    "output": 25
   },
   "run1": {
    "n": 60,
    "meanInTok": 40.617,
    "meanOutTok": 524.7,
    "truncated": 5,
    "costPerAnswerUsd": 0.013321
   },
   "run2": {
    "n": 60,
    "meanInTok": 40.617,
    "meanOutTok": 558.767,
    "truncated": 0,
    "costPerAnswerUsd": 0.014172
   }
  },
  {
   "id": "claude-opus-4-8",
   "display": "Claude Opus 4.8",
   "family": "opus",
   "listPriceUsdPerMTok": {
    "input": 5,
    "output": 25
   },
   "run1": {
    "n": 60,
    "meanInTok": 35.617,
    "meanOutTok": 522.467,
    "truncated": 3,
    "costPerAnswerUsd": 0.01324
   },
   "run2": {
    "n": 60,
    "meanInTok": 35.617,
    "meanOutTok": 550.467,
    "truncated": 0,
    "costPerAnswerUsd": 0.01394
   }
  },
  {
   "id": "claude-opus-5-5",
   "display": "Claude Opus 5.5",
   "family": "opus",
   "listPriceUsdPerMTok": {
    "input": 4,
    "output": 20
   },
   "run1": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 794.017,
    "truncated": 20,
    "costPerAnswerUsd": 0.016031
   },
   "run2": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 1065.8,
    "truncated": 2,
    "costPerAnswerUsd": 0.021466
   }
  },
  {
   "id": "claude-fable-5-1",
   "display": "Claude Fable 5.1",
   "family": "fable",
   "listPriceUsdPerMTok": {
    "input": 10,
    "output": 50
   },
   "run1": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 583.433,
    "truncated": 9,
    "costPerAnswerUsd": 0.029548
   },
   "run2": {
    "n": 60,
    "meanInTok": 37.617,
    "meanOutTok": 731.183,
    "truncated": 1,
    "costPerAnswerUsd": 0.036935
   }
  }
 ],
 "transitions": [
  {
   "from": "claude-opus-4-6",
   "to": "claude-opus-4-7",
   "label": "Claude Opus 4.6 to Claude Opus 4.7",
   "priced": true,
   "listPriceChangePct": {
    "input": 0,
    "output": 0
   },
   "outputTokensChangePct": {
    "run1": 43.302,
    "run2": 25.763
   },
   "inputTokensChangePct": 46.897,
   "costPerAnswerChangePct": {
    "run1": 43.36,
    "run2": 26.018,
    "run2Ci95": [
     6.699,
     52.2
    ]
   }
  },
  {
   "from": "claude-opus-4-7",
   "to": "claude-opus-4-8",
   "label": "Claude Opus 4.7 to Claude Opus 4.8",
   "priced": true,
   "listPriceChangePct": {
    "input": 0,
    "output": 0
   },
   "outputTokensChangePct": {
    "run1": -0.426,
    "run2": -1.485
   },
   "inputTokensChangePct": -12.31,
   "costPerAnswerChangePct": {
    "run1": -0.608,
    "run2": -1.637,
    "run2Ci95": [
     -6.307,
     3.746
    ]
   }
  },
  {
   "from": "claude-opus-4-6",
   "to": "claude-opus-5-5",
   "label": "Claude Opus 4.6 to Claude Opus 5.5",
   "priced": true,
   "listPriceChangePct": {
    "input": -20,
    "output": -20
   },
   "outputTokensChangePct": {
    "run1": 116.856,
    "run2": 139.883
   },
   "inputTokensChangePct": 36.047,
   "costPerAnswerChangePct": {
    "run1": 72.525,
    "run2": 90.877,
    "run2Ci95": [
     56.773,
     132.929
    ]
   }
  },
  {
   "from": "claude-sonnet-4-6",
   "to": "claude-sonnet-5-5",
   "label": "Claude Sonnet 4.6 to Claude Sonnet 5.5",
   "priced": true,
   "listPriceChangePct": {
    "input": -33.333,
    "output": -33.333
   },
   "outputTokensChangePct": {
    "run1": 61.074,
    "run2": 57.336
   },
   "inputTokensChangePct": 36.047,
   "costPerAnswerChangePct": {
    "run1": 7.146,
    "run2": 4.735,
    "run2Ci95": [
     -9.856,
     28.308
    ]
   }
  },
  {
   "from": "claude-haiku-4-5-20251001",
   "to": "claude-haiku-5-5",
   "label": "Claude Haiku 4.5 to Claude Haiku 5.5",
   "priced": false,
   "listPriceChangePct": null,
   "outputTokensChangePct": {
    "run1": 98.67,
    "run2": 137.465
   },
   "inputTokensChangePct": 36.047,
   "costPerAnswerChangePct": null
  }
 ],
 "battery2": {
  "prompts": 120,
  "tasks": [
   {
    "task": "code",
    "n": 30
   },
   {
    "task": "summarise",
    "n": 30
   },
   {
    "task": "long-form",
    "n": 30
   },
   {
    "task": "reasoning",
    "n": 30
   }
  ],
  "models": [
   {
    "id": "claude-haiku-4-5-20251001",
    "display": "Claude Haiku 4.5",
    "all": {
     "n": 120,
     "meanInTok": 87.958,
     "meanOutTok": 349,
     "meanThinkingTok": 0,
     "truncated": 0,
     "costPerAnswerUsd": 0.001833
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 37.8,
      "meanOutTok": 530.267,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.002689
     },
     "summarise": {
      "n": 30,
      "meanInTok": 230.433,
      "meanOutTok": 134.833,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.000905
     },
     "long-form": {
      "n": 30,
      "meanInTok": 36.2,
      "meanOutTok": 535.967,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.002716
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 47.4,
      "meanOutTok": 194.933,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.001022
     }
    }
   },
   {
    "id": "claude-haiku-5-5",
    "display": "Claude Haiku 5.5",
    "all": {
     "n": 120,
     "meanInTok": 118.042,
     "meanOutTok": 591.75,
     "meanThinkingTok": 136.575,
     "truncated": 0,
     "costPerAnswerUsd": null
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 49.9,
      "meanOutTok": 823.2,
      "meanThinkingTok": 195.3,
      "truncated": 0,
      "costPerAnswerUsd": null
     },
     "summarise": {
      "n": 30,
      "meanInTok": 318.867,
      "meanOutTok": 205.933,
      "meanThinkingTok": 26.233,
      "truncated": 0,
      "costPerAnswerUsd": null
     },
     "long-form": {
      "n": 30,
      "meanInTok": 45.733,
      "meanOutTok": 1058.333,
      "meanThinkingTok": 261.5,
      "truncated": 0,
      "costPerAnswerUsd": null
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 57.667,
      "meanOutTok": 279.533,
      "meanThinkingTok": 63.267,
      "truncated": 0,
      "costPerAnswerUsd": null
     }
    }
   },
   {
    "id": "claude-sonnet-4-6",
    "display": "Claude Sonnet 4.6",
    "all": {
     "n": 120,
     "meanInTok": 87.958,
     "meanOutTok": 422.583,
     "meanThinkingTok": 0,
     "truncated": 0,
     "costPerAnswerUsd": 0.006603
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 37.8,
      "meanOutTok": 796.833,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.012066
     },
     "summarise": {
      "n": 30,
      "meanInTok": 230.433,
      "meanOutTok": 143.367,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.002842
     },
     "long-form": {
      "n": 30,
      "meanInTok": 36.2,
      "meanOutTok": 529.167,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.008046
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 47.4,
      "meanOutTok": 220.967,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.003457
     }
    }
   },
   {
    "id": "claude-sonnet-5-5",
    "display": "Claude Sonnet 5.5",
    "all": {
     "n": 120,
     "meanInTok": 118.042,
     "meanOutTok": 534.275,
     "meanThinkingTok": 60.533,
     "truncated": 0,
     "costPerAnswerUsd": 0.005579
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 49.9,
      "meanOutTok": 809.533,
      "meanThinkingTok": 190.8,
      "truncated": 0,
      "costPerAnswerUsd": 0.008195
     },
     "summarise": {
      "n": 30,
      "meanInTok": 318.867,
      "meanOutTok": 219.9,
      "meanThinkingTok": 16.633,
      "truncated": 0,
      "costPerAnswerUsd": 0.002837
     },
     "long-form": {
      "n": 30,
      "meanInTok": 45.733,
      "meanOutTok": 891.3,
      "meanThinkingTok": 34.7,
      "truncated": 0,
      "costPerAnswerUsd": 0.009004
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 57.667,
      "meanOutTok": 216.367,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.002279
     }
    }
   },
   {
    "id": "claude-opus-4-6",
    "display": "Claude Opus 4.6",
    "all": {
     "n": 120,
     "meanInTok": 87.958,
     "meanOutTok": 361.8,
     "meanThinkingTok": 0,
     "truncated": 0,
     "costPerAnswerUsd": 0.009485
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 37.8,
      "meanOutTok": 537.333,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.013622
     },
     "summarise": {
      "n": 30,
      "meanInTok": 230.433,
      "meanOutTok": 150.667,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.004919
     },
     "long-form": {
      "n": 30,
      "meanInTok": 36.2,
      "meanOutTok": 569.6,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.014421
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 47.4,
      "meanOutTok": 189.6,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.004977
     }
    }
   },
   {
    "id": "claude-opus-4-7",
    "display": "Claude Opus 4.7",
    "all": {
     "n": 120,
     "meanInTok": 121.042,
     "meanOutTok": 502.5,
     "meanThinkingTok": 0,
     "truncated": 0,
     "costPerAnswerUsd": 0.013168
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 52.9,
      "meanOutTok": 682.667,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.017331
     },
     "summarise": {
      "n": 30,
      "meanInTok": 321.867,
      "meanOutTok": 183.3,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.006192
     },
     "long-form": {
      "n": 30,
      "meanInTok": 48.733,
      "meanOutTok": 810.033,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.020495
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 60.667,
      "meanOutTok": 334,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.008653
     }
    }
   },
   {
    "id": "claude-opus-4-8",
    "display": "Claude Opus 4.8",
    "all": {
     "n": 120,
     "meanInTok": 116.042,
     "meanOutTok": 499.742,
     "meanThinkingTok": 0,
     "truncated": 0,
     "costPerAnswerUsd": 0.013074
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 47.9,
      "meanOutTok": 719.367,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.018224
     },
     "summarise": {
      "n": 30,
      "meanInTok": 316.867,
      "meanOutTok": 179.667,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.006076
     },
     "long-form": {
      "n": 30,
      "meanInTok": 43.733,
      "meanOutTok": 793.167,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.020048
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 55.667,
      "meanOutTok": 306.767,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.007948
     }
    }
   },
   {
    "id": "claude-opus-5-5",
    "display": "Claude Opus 5.5",
    "all": {
     "n": 120,
     "meanInTok": 118.042,
     "meanOutTok": 628.958,
     "meanThinkingTok": 90.625,
     "truncated": 0,
     "costPerAnswerUsd": 0.013051
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 49.9,
      "meanOutTok": 903.8,
      "meanThinkingTok": 155.367,
      "truncated": 0,
      "costPerAnswerUsd": 0.018276
     },
     "summarise": {
      "n": 30,
      "meanInTok": 318.867,
      "meanOutTok": 254.567,
      "meanThinkingTok": 52.567,
      "truncated": 0,
      "costPerAnswerUsd": 0.006367
     },
     "long-form": {
      "n": 30,
      "meanInTok": 45.733,
      "meanOutTok": 962.433,
      "meanThinkingTok": 102.767,
      "truncated": 0,
      "costPerAnswerUsd": 0.019432
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 57.667,
      "meanOutTok": 395.033,
      "meanThinkingTok": 51.8,
      "truncated": 0,
      "costPerAnswerUsd": 0.008131
     }
    }
   },
   {
    "id": "claude-fable-5-1",
    "display": "Claude Fable 5.1",
    "all": {
     "n": 120,
     "meanInTok": 118.042,
     "meanOutTok": 474.242,
     "meanThinkingTok": 28.833,
     "truncated": 0,
     "costPerAnswerUsd": 0.024893
    },
    "byTask": {
     "code": {
      "n": 30,
      "meanInTok": 49.9,
      "meanOutTok": 651.1,
      "meanThinkingTok": 52.633,
      "truncated": 0,
      "costPerAnswerUsd": 0.033054
     },
     "summarise": {
      "n": 30,
      "meanInTok": 318.867,
      "meanOutTok": 209.8,
      "meanThinkingTok": 11.033,
      "truncated": 0,
      "costPerAnswerUsd": 0.013679
     },
     "long-form": {
      "n": 30,
      "meanInTok": 45.733,
      "meanOutTok": 850.7,
      "meanThinkingTok": 51.667,
      "truncated": 0,
      "costPerAnswerUsd": 0.042992
     },
     "reasoning": {
      "n": 30,
      "meanInTok": 57.667,
      "meanOutTok": 185.367,
      "meanThinkingTok": 0,
      "truncated": 0,
      "costPerAnswerUsd": 0.009845
     }
    }
   }
  ],
  "transitions": [
   {
    "from": "claude-opus-4-6",
    "to": "claude-opus-4-7",
    "label": "Claude Opus 4.6 to Claude Opus 4.7",
    "priced": true,
    "listPriceChangePct": {
     "input": 0,
     "output": 0
    },
    "outputTokensChangePct": 38.889,
    "all": {
     "costPerAnswerChangePct": 38.83,
     "ci95": [
      34.371,
      43.498
     ]
    },
    "byTask": {
     "code": {
      "costPerAnswerChangePct": 27.228,
      "ci95": [
       18.285,
       36.825
      ]
     },
     "summarise": {
      "costPerAnswerChangePct": 25.879,
      "ci95": [
       20.541,
       31.511
      ]
     },
     "long-form": {
      "costPerAnswerChangePct": 42.119,
      "ci95": [
       39.324,
       44.854
      ]
     },
     "reasoning": {
      "costPerAnswerChangePct": 73.86,
      "ci95": [
       63.603,
       84.447
      ]
     }
    }
   },
   {
    "from": "claude-opus-4-7",
    "to": "claude-opus-4-8",
    "label": "Claude Opus 4.7 to Claude Opus 4.8",
    "priced": true,
    "listPriceChangePct": {
     "input": 0,
     "output": 0
    },
    "outputTokensChangePct": -0.549,
    "all": {
     "costPerAnswerChangePct": -0.714,
     "ci95": [
      -2.783,
      1.412
     ]
    },
    "byTask": {
     "code": {
      "costPerAnswerChangePct": 5.153,
      "ci95": [
       0.368,
       10.693
      ]
     },
     "summarise": {
      "costPerAnswerChangePct": -1.873,
      "ci95": [
       -6.801,
       3.318
      ]
     },
     "long-form": {
      "costPerAnswerChangePct": -2.181,
      "ci95": [
       -3.431,
       -0.996
      ]
     },
     "reasoning": {
      "costPerAnswerChangePct": -8.147,
      "ci95": [
       -12.271,
       -3.032
      ]
     }
    }
   },
   {
    "from": "claude-opus-4-6",
    "to": "claude-opus-5-5",
    "label": "Claude Opus 4.6 to Claude Opus 5.5",
    "priced": true,
    "listPriceChangePct": {
     "input": -20,
     "output": -20
    },
    "outputTokensChangePct": 73.841,
    "all": {
     "costPerAnswerChangePct": 37.596,
     "ci95": [
      30.402,
      45.539
     ]
    },
    "byTask": {
     "code": {
      "costPerAnswerChangePct": 34.165,
      "ci95": [
       15.863,
       53.975
      ]
     },
     "summarise": {
      "costPerAnswerChangePct": 29.437,
      "ci95": [
       20.521,
       40.975
      ]
     },
     "long-form": {
      "costPerAnswerChangePct": 34.748,
      "ci95": [
       30.574,
       38.952
      ]
     },
     "reasoning": {
      "costPerAnswerChangePct": 63.372,
      "ci95": [
       51.004,
       77.662
      ]
     }
    }
   },
   {
    "from": "claude-sonnet-4-6",
    "to": "claude-sonnet-5-5",
    "label": "Claude Sonnet 4.6 to Claude Sonnet 5.5",
    "priced": true,
    "listPriceChangePct": {
     "input": -33.333,
     "output": -33.333
    },
    "outputTokensChangePct": 26.431,
    "all": {
     "costPerAnswerChangePct": -15.508,
     "ci95": [
      -21.526,
      -9.139
     ]
    },
    "byTask": {
     "code": {
      "costPerAnswerChangePct": -32.082,
      "ci95": [
       -40.914,
       -22.401
      ]
     },
     "summarise": {
      "costPerAnswerChangePct": -0.176,
      "ci95": [
       -5.9,
       5.689
      ]
     },
     "long-form": {
      "costPerAnswerChangePct": 11.907,
      "ci95": [
       8.644,
       15.283
      ]
     },
     "reasoning": {
      "costPerAnswerChangePct": -34.076,
      "ci95": [
       -40.258,
       -27.631
      ]
     }
    }
   },
   {
    "from": "claude-haiku-4-5-20251001",
    "to": "claude-haiku-5-5",
    "label": "Claude Haiku 4.5 to Claude Haiku 5.5",
    "priced": false,
    "listPriceChangePct": null,
    "outputTokensChangePct": 69.556,
    "all": null,
    "byTask": null
   }
  ]
 },
 "hiddenThinking": {
  "note": "Billed output tokens include hidden thinking tokens, reported by the provider in output_tokens_details.thinking_tokens. Measured on a separate recapture of the same 60 prompts (captures-v2) for all nine models. Older models (Haiku 4.5, Sonnet 4.6, Opus 4.6) do not report a thinking-token count; for them thinking is 0 because no thinking block was returned in any of the 60 answers.",
  "models": {
   "claude-haiku-4-5-20251001": {
    "thinkingBlocksReturned": 0,
    "reportedByApi": false,
    "n": 60,
    "meanOutTok": 318.567,
    "meanThinkingTok": 0,
    "thinkingShareOfBilledOutputPct": 0,
    "answersWithThinking": 0
   },
   "claude-haiku-5-5": {
    "thinkingBlocksReturned": 28,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 767.717,
    "meanThinkingTok": 163.717,
    "thinkingShareOfBilledOutputPct": 21.325,
    "answersWithThinking": 28
   },
   "claude-sonnet-4-6": {
    "thinkingBlocksReturned": 0,
    "reportedByApi": false,
    "n": 60,
    "meanOutTok": 502.317,
    "meanThinkingTok": 0,
    "thinkingShareOfBilledOutputPct": 0,
    "answersWithThinking": 0
   },
   "claude-sonnet-5-5": {
    "thinkingBlocksReturned": 14,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 772.133,
    "meanThinkingTok": 98.1,
    "thinkingShareOfBilledOutputPct": 12.705,
    "answersWithThinking": 14
   },
   "claude-opus-4-6": {
    "thinkingBlocksReturned": 0,
    "reportedByApi": false,
    "n": 60,
    "meanOutTok": 449.667,
    "meanThinkingTok": 0,
    "thinkingShareOfBilledOutputPct": 0,
    "answersWithThinking": 0
   },
   "claude-opus-4-7": {
    "thinkingBlocksReturned": 0,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 558.5,
    "meanThinkingTok": 0,
    "thinkingShareOfBilledOutputPct": 0,
    "answersWithThinking": 0
   },
   "claude-opus-4-8": {
    "thinkingBlocksReturned": 0,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 552.367,
    "meanThinkingTok": 0,
    "thinkingShareOfBilledOutputPct": 0,
    "answersWithThinking": 0
   },
   "claude-opus-5-5": {
    "thinkingBlocksReturned": 60,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 1050.383,
    "meanThinkingTok": 169.167,
    "thinkingShareOfBilledOutputPct": 16.105,
    "answersWithThinking": 60
   },
   "claude-fable-5-1": {
    "thinkingBlocksReturned": 13,
    "reportedByApi": true,
    "n": 60,
    "meanOutTok": 760.283,
    "meanThinkingTok": 32.867,
    "thinkingShareOfBilledOutputPct": 4.323,
    "answersWithThinking": 13
   }
  }
 },
 "decomposition": {
  "note": "Change in billed output tokens split into the tokenizer (same answer text counted by the newer tokenizer), visible writing (how much text the newer model writes, one tokenizer) and hidden thinking (share of billed output that is thinking). The three-way figure is compared with the provider-reported change.",
  "steps": [
   {
    "from": "claude-opus-4-6",
    "to": "claude-opus-4-7",
    "tokenizerEffectPct": 41.8,
    "visibleWritingEffectPct": -11.35,
    "hiddenThinkingEffectPct": 0,
    "thinkingShareOfBilledOutputInNewerModelPct": 0,
    "threeWayCombinedPct": 25.706,
    "providerReportedOutputTokenChangePct": 25.76
   },
   {
    "from": "claude-opus-4-7",
    "to": "claude-opus-4-8",
    "tokenizerEffectPct": 0,
    "visibleWritingEffectPct": -1.13,
    "hiddenThinkingEffectPct": 0,
    "thinkingShareOfBilledOutputInNewerModelPct": 0,
    "threeWayCombinedPct": -1.13,
    "providerReportedOutputTokenChangePct": -1.49
   },
   {
    "from": "claude-opus-4-6",
    "to": "claude-opus-5-5",
    "tokenizerEffectPct": 41.8,
    "visibleWritingEffectPct": 39.59,
    "hiddenThinkingEffectPct": 19.197,
    "thinkingShareOfBilledOutputInNewerModelPct": 16.105,
    "threeWayCombinedPct": 135.936,
    "providerReportedOutputTokenChangePct": 139.88
   },
   {
    "from": "claude-sonnet-4-6",
    "to": "claude-sonnet-5-5",
    "tokenizerEffectPct": 42.76,
    "visibleWritingEffectPct": -1.66,
    "hiddenThinkingEffectPct": 14.554,
    "thinkingShareOfBilledOutputInNewerModelPct": 12.705,
    "threeWayCombinedPct": 60.823,
    "providerReportedOutputTokenChangePct": 57.34
   },
   {
    "from": "claude-haiku-4-5-20251001",
    "to": "claude-haiku-5-5",
    "tokenizerEffectPct": 45.08,
    "visibleWritingEffectPct": 28.77,
    "hiddenThinkingEffectPct": 27.105,
    "thinkingShareOfBilledOutputInNewerModelPct": 21.325,
    "threeWayCombinedPct": 137.457,
    "providerReportedOutputTokenChangePct": 137.46
   }
  ]
 },
 "runToRunNoise": {
  "note": "No model was free of cut-offs in both runs, so there is no clean run-to-run noise estimate. Uncertainty is shown by the paired bootstrap interval over prompts instead.",
  "models": []
 },
 "battery": {
  "prompts": 60,
  "source": "research/ai-model-degradation-index-2026/battery.jsonl (frozen)"
 }
}