{
  "schemaVersion": "1.2.0",
  "resultId": "builderbench-2026-10-08-luna-haiku",
  "slug": "2026-10-08-luna-haiku",
  "title": "BuilderBench v3: GPT-6 Luna and Haiku 5.5",
  "runDate": "2026-10-08",
  "displayDate": "8 October 2026",
  "timezone": "Australia/Perth",
  "status": "cleared_for_web",
  "benchmarkVersion": "v3",
  "packVersion": "3.0.0",
  "packSha256": "37e2da6c5769143b80027b2be176930fd453c5a4b7499001a3692bd8ee664251",
  "methodologyVersion": "v3",
  "scope": "One owner-finalized partial personal assessment of one GPT-6 Luna run with Codex and one Haiku 5.5 run with Claude on the v3 pack. This is not a universal model ranking or a statistical claim.",
  "durationSeconds": 14400,
  "scoringWeights": {
    "quality": 70,
    "autonomy": 20,
    "proactivity": 10
  },
  "settings": {
    "reasoning": "Extra High (xhigh)",
    "speed": "Standard",
    "billing": "Existing subscriptions (text routes)"
  },
  "review": {
    "reviewer": "Lennox Saint",
    "sampleSizePerSetup": 1,
    "scoresLocked": true,
    "identitiesRevealed": true,
    "scoreDigestSha256": "cf0497192edc628280f1deb6756e335cbaf6fab9f2eb7361b3f488e05c55f31b",
    "assessment": "Owner-finalized partial personal assessment",
    "conditions": {
      "mode": "blind",
      "detail": "Requested as a blind comparison. The build work was rated as Candidate A and Candidate B, and the names were revealed after score lock. It was not perfectly blind: earlier operator diagnostic output included cost clues. The four native sessions were rated by model name in a separate form that never shows which candidate is which."
    }
  },
  "candidates": [
    {
      "name": "GPT-6 Luna",
      "harness": "Codex",
      "modelVersion": null,
      "harnessVersion": null,
      "quality": 45.12,
      "autonomy": 40,
      "proactivity": 50,
      "overall": 44.58,
      "tasks": [
        {
          "label": "Writing",
          "score": 18,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.02720112,
            "coverage": "recorded"
          }
        },
        {
          "label": "Landing page",
          "score": 30,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.13938974,
            "coverage": "recorded"
          }
        },
        {
          "label": "App lead magnet",
          "score": 50,
          "weight": 20,
          "cost": {
            "apiEquivalentEstimateUsd": 0.22447052,
            "coverage": "recorded"
          }
        },
        {
          "label": "Video editing",
          "score": 11.67,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.48432552,
            "coverage": "recorded"
          }
        },
        {
          "label": "Motion",
          "score": 15,
          "weight": 5,
          "cost": {
            "apiEquivalentEstimateUsd": 0.41752644,
            "coverage": "recorded"
          }
        },
        {
          "label": "Thumbnails",
          "score": 12,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.10449346,
            "coverage": "recorded"
          }
        },
        {
          "label": "Browser + computer",
          "score": 55,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": null,
            "coverage": "unavailable"
          }
        },
        {
          "label": "Research",
          "score": 84.5,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.0291092,
            "coverage": "recorded"
          }
        },
        {
          "label": "Automation",
          "score": 100,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 0.04650618,
            "coverage": "recorded"
          }
        },
        {
          "label": "Founder conversations",
          "score": 65,
          "weight": 5,
          "cost": {
            "apiEquivalentEstimateUsd": null,
            "coverage": "unavailable"
          }
        }
      ],
      "cost": {
        "actualIncrementalSpendUsd": null,
        "apiEquivalentEstimateUsd": 1.47302218,
        "coverage": "Partial API-equivalent estimate: 18 of 18 saved execution records priced with a dated Standard short-context token scenario. Native sessions, unknown cache writes and context premiums are not included.",
        "pricedRecords": {
          "priced": 18,
          "total": 18
        },
        "estimate": {
          "status": "partial",
          "method": "Standard short-context token scenario: $0.10 input, $0.01 cached input and $0.50 output per 1M tokens.",
          "pricedAt": "2026-10-08"
        }
      },
      "apiPricing": {
        "status": "verified",
        "unit": "per 1M tokens",
        "processing": "Standard",
        "checkedAt": "2026-10-08",
        "sourceUrl": "https://developers.openai.com/api/docs/models/gpt-6-luna",
        "sourceLabel": "OpenAI API docs: GPT-6 Luna",
        "tiers": [
          {
            "label": "Up to 272K input",
            "inputUsd": 0.1,
            "cacheReadUsd": 0.01,
            "outputUsd": 0.5
          },
          {
            "label": "Above 272K input",
            "inputUsd": 0.2,
            "cacheReadUsd": 0.02,
            "outputUsd": 0.75
          }
        ],
        "notes": "Cache writes: $0.125 / $0.25. Tool and regional fees are separate. The run estimate uses the short-context row; request-level bands and cache writes were unavailable."
      },
      "nativeRatings": [
        {
          "lane": "browser",
          "score": 50
        },
        {
          "lane": "computer",
          "score": 60
        },
        {
          "lane": "advice",
          "score": 80
        },
        {
          "lane": "planning",
          "score": 50
        }
      ]
    },
    {
      "name": "Haiku 5.5",
      "harness": "Claude",
      "modelVersion": null,
      "harnessVersion": null,
      "quality": 43.11,
      "autonomy": 30,
      "proactivity": 30,
      "overall": 39.18,
      "tasks": [
        {
          "label": "Writing",
          "score": 21,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 1.3086604,
            "coverage": "recorded"
          }
        },
        {
          "label": "Landing page",
          "score": 30,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 6.1069003,
            "coverage": "recorded"
          }
        },
        {
          "label": "App lead magnet",
          "score": 48.75,
          "weight": 20,
          "cost": {
            "apiEquivalentEstimateUsd": 5.0375584,
            "coverage": "partial"
          }
        },
        {
          "label": "Video editing",
          "score": 5,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 9.1086284,
            "coverage": "recorded"
          }
        },
        {
          "label": "Motion",
          "score": 10,
          "weight": 5,
          "cost": {
            "apiEquivalentEstimateUsd": 7.2435023,
            "coverage": "recorded"
          }
        },
        {
          "label": "Thumbnails",
          "score": 26,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 4.577406,
            "coverage": "recorded"
          }
        },
        {
          "label": "Browser + computer",
          "score": 55,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": null,
            "coverage": "unavailable"
          }
        },
        {
          "label": "Research",
          "score": 82,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 1.416462,
            "coverage": "recorded"
          }
        },
        {
          "label": "Automation",
          "score": 83.33,
          "weight": 10,
          "cost": {
            "apiEquivalentEstimateUsd": 2.5922094,
            "coverage": "recorded"
          }
        },
        {
          "label": "Founder conversations",
          "score": 52.5,
          "weight": 5,
          "cost": {
            "apiEquivalentEstimateUsd": null,
            "coverage": "unavailable"
          }
        }
      ],
      "cost": {
        "actualIncrementalSpendUsd": null,
        "apiEquivalentEstimateUsd": 37.3913272,
        "coverage": "Partial API-equivalent estimate: 20 of 21 saved execution records priced, keeping provider-reported estimates including recorded tools. Failed and native-session usage is unknown.",
        "pricedRecords": {
          "priced": 20,
          "total": 21
        },
        "estimate": {
          "status": "partial",
          "method": "Provider-reported estimates, including recorded tools.",
          "pricedAt": "2026-10-08"
        }
      },
      "apiPricing": {
        "status": "verified",
        "unit": "per 1M tokens",
        "processing": "Standard",
        "checkedAt": "2026-10-08",
        "sourceUrl": "https://platform.claude.com/docs/en/models/haiku-5-5/overview",
        "sourceLabel": "Claude API docs: Haiku 5.5",
        "tiers": [
          {
            "label": "Up to 100K input",
            "inputUsd": 0.1,
            "cacheReadUsd": 0.01,
            "outputUsd": 0.5
          },
          {
            "label": "Above 100K input",
            "inputUsd": 0.5,
            "cacheReadUsd": 0.05,
            "outputUsd": 2.5
          }
        ],
        "notes": "5-minute cache writes: $0.125 / $0.625. 1-hour cache writes: $0.20 / $1.00. Tool fees are separate. The run estimate keeps provider-reported figures instead of applying these rates."
      },
      "nativeRatings": [
        {
          "lane": "browser",
          "score": 50
        },
        {
          "lane": "computer",
          "score": 60
        },
        {
          "lane": "advice",
          "score": 60
        },
        {
          "lane": "planning",
          "score": 45
        }
      ]
    }
  ],
  "costNote": "API-equivalent estimates price recorded execution usage only. They are not subscription bills. GPT-6 Luna uses a dated Standard short-context token scenario. Haiku 5.5 keeps provider-reported estimates, including recorded tools. Both are partial and use different methods, so they do not support an exact cost ratio. Native sessions, unknown cache writes, context premiums, regional fees, setup, calibration, coordination, grading, presentation and subscription fees are excluded. Actual incremental spend is unknown.",
  "methodologyNote": "The owner rated the saved build outputs, the four native sessions, autonomy and proactivity. Task 9 technical repair is conductor-scored with recorded evidence. For this run only, the owner also delegated grading of the research memo and decision table to the conductor. The lead-magnet app keeps this run's original frozen four-part rubric. The saved scores were finalized as a partial personal assessment and locked. Quality is the weighted task score. v1 to v3 use 70% quality, 20% autonomy and 10% proactivity.",
  "limitations": [
    "This is one owner-finalized partial personal assessment of one run per setup. It is not a universal ranking and it is not statistically conclusive.",
    "A 5.40-point gap from one run per setup does not show that one model is better in general.",
    "The original execution is incomplete. The launcher stopped just before the deadline, and the saved outputs were collected without a rerun.",
    "One Haiku 5.5 app revision ended with an authentication failure. Its saved output was assessed as delivered.",
    "Native browser and computer recordings are incomplete. Those two lanes carry owner ratings without accepted continuous recordings.",
    "Video, motion and thumbnail files passed technical decoding and sampled visual checks only. Audio and continuous viewing were not independently verified.",
    "The review was not perfectly blind. Earlier operator diagnostic output included cost clues before the scores were locked.",
    "The founder conversations were archived after score lock, without session timing or equal-context proof.",
    "Research memo and decision-table grading was delegated to the conductor for this run only.",
    "The two cost estimates are partial and use different methods. They do not support an exact cost ratio.",
    "Actual incremental spend is unknown because the run used existing subscriptions.",
    "The score belongs to the full setup: model, harness, tools and settings.",
    "Pack 3.0.0 scores compare only with other pack 3.0.0 results, not with v1 or v2 results.",
    "The build window was four hours. Native live sessions were timed separately from the build window."
  ],
  "assets": {
    "resultGraphic": {
      "src": "/builderbench/2026-10-08-luna-haiku/luna-haiku-reveal-v2.png",
      "sha256": "3aefae4b2be12097104402303e64ccb027259e941dbbf9399f4c5b96b1e85e10",
      "width": 1672,
      "height": 941,
      "alt": "BuilderBench by the LAB, the reveal: GPT-6 Luna scored 44.58 out of 100 and Haiku 5.5 scored 39.18 out of 100. Luna leads by 5.40 points. Final personal scores with partial evidence, 8 October 2026, v3, four hours, Extra High. API estimates: Luna US$1.47 from a short-context scenario and Haiku US$37.39 provider-reported, both partial recorded usage and not bills. Base rates per 1M tokens for both: $0.10 input, $0.01 cache read, $0.50 output."
    }
  },
  "sourceResultsSha256": "d8fed2b545ec4954161b58136f5f05bc1afb5d1b8e01f97c2dd1b41f348e6407",
  "integrity": {
    "algorithm": "sha256",
    "digest": "d5cf1ad99af630a981bfae255aa035b8ca2a768cd0f00afa00722bf71b658e72"
  }
}