{
  "schemaVersion": 1,
  "runs": [
    {
      "id": "2026-10-27-windows-gemma4-12b-native-0-1-54",
      "provenance": {
        "startedAt": "2026-10-28T07:03:41.388Z",
        "device": {
          "label": "win32 · AMD Ryzen AI 9 HX 370 w/ Radeon 890M",
          "platform": "win32",
          "arch": "x64",
          "memoryGb": 31,
          "osRelease": "win32 10.0.26300",
          "cpuModel": "AMD Ryzen AI 9 HX 370 w/ Radeon 890M",
          "gpuModel": "NVIDIA GeForce RTX 5070 Ti Laptop GPU"
        },
        "harnessCommit": "22fec99111a5ad9796af83a4a1a222236be91006",
        "sourceHash": "10a41de112c12aa33f12715ea0e1f30666cce2a8bf657064944820ed9f92b2e4",
        "serviceBuildHash": "8902bc3d7cf1ce309fbe752cdbff9a9a3dea0b27ef589f884c07c41e21602ace",
        "catalogHash": "2f12c2abbb9c4b216cda840c89d5683e7798b35e27fd683ea1f9bd864aa93e19",
        "gildeVersion": "0.3.1",
        "count": 3,
        "judgeModelId": "codex-inline/current-session",
        "engineBuilds": {
          "llama-cpp": "CUDA 12.8 / 620d2225cc436ab0b001e3200e590e3b286ef9ab / native 0.1.54"
        }
      },
      "scope": "suite",
      "memoryCoverage": "system-only",
      "suites": [
        "core",
        "productivity",
        "developer",
        "complex-work"
      ],
      "scenariosBySuite": {
        "core": [
          "tictactoe",
          "petshop",
          "tankcombat",
          "schema-migration",
          "failing-tests-spec",
          "symptom-debug",
          "data-wrangle",
          "incident-postmortem",
          "ops-runbook-anomaly",
          "plan-and-estimate",
          "conflict-synthesis"
        ],
        "productivity": [
          "constrained-comms",
          "records-intake",
          "plan-and-estimate",
          "meeting-followup",
          "craftbook-spreadsheet-model",
          "conflict-synthesis",
          "docblocks-theme-roundtrip",
          "craftbook-powerpoint-deck",
          "wikipedia-research-brief",
          "craftbook-week-plan",
          "craftbook-ab-test-readout",
          "craftbook-annotated-bibliography",
          "craftbook-research-to-document"
        ],
        "developer": [
          "fix-squisq-bugs",
          "dev-craftbook-routing",
          "interface-contract",
          "craftbook-code-review",
          "craftbook-api-contract-review",
          "craftbook-deep-security-review",
          "large-pr-review",
          "craftbook-codemod-sweep",
          "craftbook-refactor-module",
          "craftbook-bug-fix-tdd"
        ],
        "complex-work": [
          "craftbook-find-vs-create",
          "craftbook-edit-midtask",
          "craftbook-invoice-run",
          "craftbook-author-linear",
          "craftbook-route-multi",
          "craftbook-export-generalize",
          "craftbook-author-params",
          "craftbook-author-gate-script",
          "craftbook-author-fanout"
        ]
      },
      "note": "All four suites completed: 129 attempts, exactly three per suite/scenario, 89 raw passes. The eligible subset is 89/116 after 12 required-tool exclusions and one unevaluable authored gate; that last defect was model-authored, not a machine outage. All attempts remain in task performance. There are 59 artifact judgments: 56 included and three excluded for a mismatched A/B rubric, plus 70 tasks without an applicable rubric. Different rubrics are kept separate. Nine deferred canonical browser checks passed on the unchanged saved artifacts. Reviews assessed source and artifacts, not rendered visual polish. Admission-probe speeds remain separate from task medians. See the October 28 Windows Gemma 4 report download for exact axes, rationales, raw scores, exclusions, tokens, GPU telemetry and provenance."
    }
  ],
  "results": [
    {
      "modelId": "gemma4-12b-q4",
      "label": "Gemma 4 (12B, Q4)",
      "engine": "llama-cpp",
      "tier": "medium",
      "parameterSize": "12B",
      "quantization": "UD-Q4_K_XL",
      "runId": "2026-10-27-windows-gemma4-12b-native-0-1-54",
      "suiteId": "core",
      "performance": {
        "prefillTokensPerSec": 1754,
        "decodeTokensPerSec": 140.9,
        "samples": 1
      },
      "runtime": {
        "contextTokens": 152576,
        "peakMemoryMb": 14531,
        "kvCacheType": "f16"
      },
      "taskMeasurements": [
        {
          "id": "complete-run",
          "label": "All trials · native 0.1.54",
          "trials": 33,
          "source": "engine-log",
          "outputTokensPerSec": {
            "samples": 33,
            "median": 126.21,
            "min": 99.26,
            "max": 148.97
          },
          "prefillTokensPerSec": {
            "samples": 33,
            "median": 1927.35,
            "min": 1646.2,
            "max": 2022.87
          },
          "durationMs": {
            "samples": 33,
            "median": 57117,
            "min": 29954,
            "max": 726372
          },
          "peakRssMb": {
            "samples": 33,
            "median": 14531,
            "min": 10754,
            "max": 21156
          },
          "peakGpuMemMb": {
            "samples": 33,
            "median": 10606,
            "min": 10603,
            "max": 10648
          },
          "judgments": [
            {
              "meanScore": 4.527777777777778,
              "artifacts": 9,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "default-html-v1"
            }
          ],
          "noArtifact": 0,
          "excludedReviews": 0,
          "notApplicableReviews": 24
        }
      ],
      "cells": [
        {
          "scenarioId": "tictactoe",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 49390
        },
        {
          "scenarioId": "petshop",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 147724
        },
        {
          "scenarioId": "tankcombat",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 56874
        },
        {
          "scenarioId": "schema-migration",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 162720
        },
        {
          "scenarioId": "failing-tests-spec",
          "trials": 3,
          "successes": 2,
          "nonModelFailures": 0,
          "medianDurationMs": 31136
        },
        {
          "scenarioId": "symptom-debug",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 43281
        },
        {
          "scenarioId": "data-wrangle",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 0,
          "medianDurationMs": 555321
        },
        {
          "scenarioId": "incident-postmortem",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 113453
        },
        {
          "scenarioId": "ops-runbook-anomaly",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 74794
        },
        {
          "scenarioId": "plan-and-estimate",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 30130
        },
        {
          "scenarioId": "conflict-synthesis",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 35771
        }
      ]
    },
    {
      "modelId": "gemma4-12b-q4",
      "label": "Gemma 4 (12B, Q4)",
      "engine": "llama-cpp",
      "tier": "medium",
      "parameterSize": "12B",
      "quantization": "UD-Q4_K_XL",
      "runId": "2026-10-27-windows-gemma4-12b-native-0-1-54",
      "suiteId": "productivity",
      "performance": {
        "prefillTokensPerSec": 1754,
        "decodeTokensPerSec": 140.9,
        "samples": 1
      },
      "runtime": {
        "contextTokens": 152576,
        "peakMemoryMb": 18272,
        "kvCacheType": "f16"
      },
      "taskMeasurements": [
        {
          "id": "complete-run",
          "label": "All trials · native 0.1.54",
          "trials": 39,
          "source": "engine-log",
          "outputTokensPerSec": {
            "samples": 39,
            "median": 122.51,
            "min": 100.95,
            "max": 143.1
          },
          "prefillTokensPerSec": {
            "samples": 39,
            "median": 1857.89,
            "min": 1733.81,
            "max": 2087.06
          },
          "durationMs": {
            "samples": 39,
            "median": 153932,
            "min": 29051,
            "max": 4778004
          },
          "peakRssMb": {
            "samples": 39,
            "median": 18272,
            "min": 10750,
            "max": 21590
          },
          "peakGpuMemMb": {
            "samples": 39,
            "median": 10606,
            "min": 10600,
            "max": 10606
          },
          "judgments": [
            {
              "meanScore": 6.416666666666667,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-annotated-bibliography"
            },
            {
              "meanScore": 6.583333333333333,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-powerpoint-deck"
            },
            {
              "meanScore": 6.111111111111111,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-research-to-document"
            },
            {
              "meanScore": 4.333333333333333,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-spreadsheet-model"
            },
            {
              "meanScore": 1.25,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-week-plan"
            },
            {
              "meanScore": 5.333333333333333,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "docblocks-theme-roundtrip"
            },
            {
              "meanScore": 6.416666666666667,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "meeting-followup"
            },
            {
              "meanScore": 4.25,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "wikipedia-research-brief"
            }
          ],
          "noArtifact": 0,
          "excludedReviews": 3,
          "notApplicableReviews": 12
        }
      ],
      "cells": [
        {
          "scenarioId": "constrained-comms",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 29096
        },
        {
          "scenarioId": "records-intake",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 129980
        },
        {
          "scenarioId": "plan-and-estimate",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 30185
        },
        {
          "scenarioId": "meeting-followup",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 169601
        },
        {
          "scenarioId": "craftbook-spreadsheet-model",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 147432
        },
        {
          "scenarioId": "conflict-synthesis",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 91504
        },
        {
          "scenarioId": "docblocks-theme-roundtrip",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 3,
          "medianDurationMs": 512110
        },
        {
          "scenarioId": "craftbook-powerpoint-deck",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 153932
        },
        {
          "scenarioId": "wikipedia-research-brief",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 186311
        },
        {
          "scenarioId": "craftbook-week-plan",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 495743
        },
        {
          "scenarioId": "craftbook-ab-test-readout",
          "trials": 3,
          "successes": 2,
          "nonModelFailures": 0,
          "medianDurationMs": 291565
        },
        {
          "scenarioId": "craftbook-annotated-bibliography",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 167769
        },
        {
          "scenarioId": "craftbook-research-to-document",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 173220
        }
      ]
    },
    {
      "modelId": "gemma4-12b-q4",
      "label": "Gemma 4 (12B, Q4)",
      "engine": "llama-cpp",
      "tier": "medium",
      "parameterSize": "12B",
      "quantization": "UD-Q4_K_XL",
      "runId": "2026-10-27-windows-gemma4-12b-native-0-1-54",
      "suiteId": "developer",
      "performance": {
        "prefillTokensPerSec": 1754,
        "decodeTokensPerSec": 140.9,
        "samples": 1
      },
      "runtime": {
        "contextTokens": 152576,
        "peakMemoryMb": 18904,
        "kvCacheType": "f16"
      },
      "taskMeasurements": [
        {
          "id": "complete-run",
          "label": "All trials · native 0.1.54",
          "trials": 30,
          "source": "engine-log",
          "outputTokensPerSec": {
            "samples": 30,
            "median": 121.85,
            "min": 109.44,
            "max": 139.43
          },
          "prefillTokensPerSec": {
            "samples": 30,
            "median": 1738.525,
            "min": 1236.6,
            "max": 1947.13
          },
          "durationMs": {
            "samples": 30,
            "median": 313359,
            "min": 49762,
            "max": 2653762
          },
          "peakRssMb": {
            "samples": 30,
            "median": 18904,
            "min": 14100,
            "max": 21846
          },
          "peakGpuMemMb": {
            "samples": 30,
            "median": 10606,
            "min": 10515,
            "max": 10730
          },
          "judgments": [
            {
              "meanScore": 5,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-api-contract-review"
            },
            {
              "meanScore": 4.222222222222222,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-bug-fix-tdd"
            },
            {
              "meanScore": 7.166666666666667,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-code-review"
            },
            {
              "meanScore": 3.4444444444444446,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-codemod-sweep"
            },
            {
              "meanScore": 5.111111111111111,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-deep-security-review"
            },
            {
              "meanScore": 1.8888888888888886,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-refactor-module"
            }
          ],
          "noArtifact": 0,
          "excludedReviews": 0,
          "notApplicableReviews": 12
        }
      ],
      "cells": [
        {
          "scenarioId": "fix-squisq-bugs",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 159230
        },
        {
          "scenarioId": "dev-craftbook-routing",
          "trials": 3,
          "successes": 1,
          "nonModelFailures": 0,
          "medianDurationMs": 274213
        },
        {
          "scenarioId": "interface-contract",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 52574
        },
        {
          "scenarioId": "craftbook-code-review",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 114695
        },
        {
          "scenarioId": "craftbook-api-contract-review",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 302320
        },
        {
          "scenarioId": "craftbook-deep-security-review",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 0,
          "medianDurationMs": 175311
        },
        {
          "scenarioId": "large-pr-review",
          "trials": 3,
          "successes": 2,
          "nonModelFailures": 0,
          "medianDurationMs": 1349609
        },
        {
          "scenarioId": "craftbook-codemod-sweep",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 3,
          "medianDurationMs": 984659
        },
        {
          "scenarioId": "craftbook-refactor-module",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 3,
          "medianDurationMs": 804181
        },
        {
          "scenarioId": "craftbook-bug-fix-tdd",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 3,
          "medianDurationMs": 1727753
        }
      ]
    },
    {
      "modelId": "gemma4-12b-q4",
      "label": "Gemma 4 (12B, Q4)",
      "engine": "llama-cpp",
      "tier": "medium",
      "parameterSize": "12B",
      "quantization": "UD-Q4_K_XL",
      "runId": "2026-10-27-windows-gemma4-12b-native-0-1-54",
      "suiteId": "complex-work",
      "performance": {
        "prefillTokensPerSec": 1754,
        "decodeTokensPerSec": 140.9,
        "samples": 1
      },
      "runtime": {
        "contextTokens": 157696,
        "peakMemoryMb": 19753,
        "kvCacheType": "f16"
      },
      "taskMeasurements": [
        {
          "id": "complete-run",
          "label": "All trials · native 0.1.54",
          "trials": 27,
          "source": "engine-log",
          "outputTokensPerSec": {
            "samples": 27,
            "median": 108.41,
            "min": 97.1,
            "max": 142.73
          },
          "prefillTokensPerSec": {
            "samples": 27,
            "median": 1731.36,
            "min": 1259.77,
            "max": 1869.46
          },
          "durationMs": {
            "samples": 27,
            "median": 389416,
            "min": 48152,
            "max": 2585107
          },
          "peakRssMb": {
            "samples": 27,
            "median": 19753,
            "min": 12187,
            "max": 21991
          },
          "peakGpuMemMb": {
            "samples": 27,
            "median": 10515,
            "min": 10515,
            "max": 10521
          },
          "judgments": [
            {
              "meanScore": 5.533333333333334,
              "artifacts": 3,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "craftbook-invoice-run"
            },
            {
              "meanScore": 2.625,
              "artifacts": 2,
              "judgeModel": "codex-inline/current-session",
              "rubricId": "default-html-v1"
            }
          ],
          "noArtifact": 0,
          "excludedReviews": 0,
          "notApplicableReviews": 22
        }
      ],
      "cells": [
        {
          "scenarioId": "craftbook-find-vs-create",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 49091
        },
        {
          "scenarioId": "craftbook-edit-midtask",
          "trials": 3,
          "successes": 2,
          "nonModelFailures": 0,
          "medianDurationMs": 172467
        },
        {
          "scenarioId": "craftbook-invoice-run",
          "trials": 3,
          "successes": 3,
          "nonModelFailures": 0,
          "medianDurationMs": 192438
        },
        {
          "scenarioId": "craftbook-author-linear",
          "trials": 3,
          "successes": 1,
          "nonModelFailures": 0,
          "medianDurationMs": 389416
        },
        {
          "scenarioId": "craftbook-route-multi",
          "trials": 3,
          "successes": 1,
          "nonModelFailures": 0,
          "medianDurationMs": 343338
        },
        {
          "scenarioId": "craftbook-export-generalize",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 0,
          "medianDurationMs": 761753
        },
        {
          "scenarioId": "craftbook-author-params",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 0,
          "medianDurationMs": 448496
        },
        {
          "scenarioId": "craftbook-author-gate-script",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 1,
          "medianDurationMs": 635005
        },
        {
          "scenarioId": "craftbook-author-fanout",
          "trials": 3,
          "successes": 0,
          "nonModelFailures": 0,
          "medianDurationMs": 721741
        }
      ]
    }
  ]
}
