{
  "generated_date": "2026-10-28",
  "model_names": {
    "4B-Q4": "Qwen3.5-4B Q4_K_M",
    "9B-Q4": "Qwen3.5-9B Q4_K_M",
    "9B-Q6": "Qwen3.5-9B Q6_K",
    "27B-Q4": "Qwen3.8-27B UD-Q4_K_M",
    "35B-A3B": "Qwen3.6-35B-A3B UD-Q4_K_XL / Strata",
    "Flash IQ2": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
    "Flash XXS": "Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS / Strata",
    "Flash S": "Qwen3.8-Flash-Next GSQ-RCO IQ3_S / Strata"
  },
  "formula": "(2*V + T) / (2*criteria)",
  "H_definition": "Additional unsupported claims outside selected criteria; not a complete hallucination rate.",
  "gp60": [
    {
      "corpus": "GP60",
      "cohort": "R09",
      "source_key": "4b_q4",
      "model": "4B-Q4",
      "full_name": "Qwen3.5-4B Q4_K_M",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 416,
      "T": 158,
      "A": 128,
      "F": 18,
      "F_priority": 12,
      "H_unsupported_claims": 15,
      "points": 990,
      "max_points": 1440,
      "weighted_percent": 68.75,
      "fully_correct_percent": 57.77777777777777,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "resident-gp60-2026-10-27-09/summary/model_summary.csv",
      "criterion_source": "resident-gp60-2026-10-27-09/summary/fact_reviews.csv",
      "id": "01"
    },
    {
      "corpus": "GP60",
      "cohort": "R09",
      "source_key": "9b_q4",
      "model": "9B-Q4",
      "full_name": "Qwen3.5-9B Q4_K_M",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 528,
      "T": 116,
      "A": 58,
      "F": 18,
      "F_priority": 13,
      "H_unsupported_claims": 8,
      "points": 1172,
      "max_points": 1440,
      "weighted_percent": 81.38888888888889,
      "fully_correct_percent": 73.33333333333333,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "resident-gp60-2026-10-27-09/summary/model_summary.csv",
      "criterion_source": "resident-gp60-2026-10-27-09/summary/fact_reviews.csv",
      "id": "02"
    },
    {
      "corpus": "GP60",
      "cohort": "R09",
      "source_key": "9b_q6",
      "model": "9B-Q6",
      "full_name": "Qwen3.5-9B Q6_K",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 541,
      "T": 104,
      "A": 53,
      "F": 22,
      "F_priority": 14,
      "H_unsupported_claims": 6,
      "points": 1186,
      "max_points": 1440,
      "weighted_percent": 82.36111111111111,
      "fully_correct_percent": 75.13888888888889,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "resident-gp60-2026-10-27-09/summary/model_summary.csv",
      "criterion_source": "resident-gp60-2026-10-27-09/summary/fact_reviews.csv",
      "id": "03"
    },
    {
      "corpus": "GP60",
      "cohort": "R09",
      "source_key": "27b_q4",
      "model": "27B-Q4",
      "full_name": "Qwen3.8-27B UD-Q4_K_M",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 638,
      "T": 63,
      "A": 17,
      "F": 2,
      "F_priority": 0,
      "H_unsupported_claims": 17,
      "points": 1339,
      "max_points": 1440,
      "weighted_percent": 92.98611111111111,
      "fully_correct_percent": 88.61111111111111,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "resident-gp60-2026-10-27-09/summary/model_summary.csv",
      "criterion_source": "resident-gp60-2026-10-27-09/summary/fact_reviews.csv",
      "id": "04"
    },
    {
      "corpus": "GP60",
      "cohort": "R12",
      "source_key": "27b_q4",
      "model": "27B-Q4",
      "full_name": "Qwen3.8-27B UD-Q4_K_M",
      "profile": "32k · 1 Slot · Neubewertung",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 620,
      "T": 85,
      "A": 14,
      "F": 1,
      "F_priority": 0,
      "H_unsupported_claims": 20,
      "points": 1325,
      "max_points": 1440,
      "weighted_percent": 92.01388888888889,
      "fully_correct_percent": 86.11111111111111,
      "schema_ok": 60,
      "note": "Dieselben Antworten wie R09; erneut maskiert bewertet.",
      "same_answers_as": "R09/27b_q4",
      "summary_source": "strata-iq3-2026-10-27-12/summary/comparison.csv",
      "criterion_source": "strata-iq3-2026-10-27-12/summary/fact_reviews.csv",
      "id": "05"
    },
    {
      "corpus": "GP60",
      "cohort": "R14",
      "source_key": "qwen36",
      "model": "35B-A3B",
      "full_name": "Qwen3.6-35B-A3B UD-Q4_K_XL / Strata",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 526,
      "T": 137,
      "A": 41,
      "F": 16,
      "F_priority": 9,
      "H_unsupported_claims": 22,
      "points": 1189,
      "max_points": 1440,
      "weighted_percent": 82.56944444444444,
      "fully_correct_percent": 73.05555555555556,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-qwen36-iq3s24-2026-10-28-14/summary/results.json",
      "criterion_source": "strata-qwen36-iq3s24-2026-10-28-14/summary/fact_reviews.csv",
      "id": "06"
    },
    {
      "corpus": "GP60",
      "cohort": "R11",
      "source_key": "minimum",
      "model": "Flash IQ2",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
      "profile": "512 MiB Cache",
      "cache_mib": 512,
      "cases": 60,
      "criteria": 720,
      "V": 602,
      "T": 98,
      "A": 15,
      "F": 5,
      "F_priority": 2,
      "H_unsupported_claims": 6,
      "points": 1302,
      "max_points": 1440,
      "weighted_percent": 90.41666666666667,
      "fully_correct_percent": 83.61111111111111,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-mincache-2026-10-27-11/summary/results.json",
      "criterion_source": "strata-mincache-2026-10-27-11/summary/fact_reviews.csv",
      "id": "07"
    },
    {
      "corpus": "GP60",
      "cohort": "R12",
      "source_key": "IQ2_XS",
      "model": "Flash IQ2",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
      "profile": "512 MiB Cache",
      "cache_mib": 512,
      "cases": 60,
      "criteria": 720,
      "V": 585,
      "T": 114,
      "A": 11,
      "F": 10,
      "F_priority": 5,
      "H_unsupported_claims": 6,
      "points": 1284,
      "max_points": 1440,
      "weighted_percent": 89.16666666666667,
      "fully_correct_percent": 81.25,
      "schema_ok": 59,
      "note": "Eine Antwort verletzt das JSON-Schema.",
      "same_answers_as": null,
      "summary_source": "strata-iq3-2026-10-27-12/summary/comparison.csv",
      "criterion_source": "strata-iq3-2026-10-27-12/summary/fact_reviews.csv",
      "id": "08"
    },
    {
      "corpus": "GP60",
      "cohort": "R09",
      "source_key": "flash_iq2xs",
      "model": "Flash IQ2",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
      "profile": "groß · 12-GiB-Prozessziel",
      "cache_mib": null,
      "cases": 60,
      "criteria": 720,
      "V": 586,
      "T": 105,
      "A": 21,
      "F": 8,
      "F_priority": 5,
      "H_unsupported_claims": 8,
      "points": 1277,
      "max_points": 1440,
      "weighted_percent": 88.68055555555556,
      "fully_correct_percent": 81.38888888888889,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "resident-gp60-2026-10-27-09/summary/model_summary.csv",
      "criterion_source": "resident-gp60-2026-10-27-09/summary/fact_reviews.csv",
      "id": "09"
    },
    {
      "corpus": "GP60",
      "cohort": "R11",
      "source_key": "full",
      "model": "Flash IQ2",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
      "profile": "6196 MiB Cache",
      "cache_mib": 6196,
      "cases": 60,
      "criteria": 720,
      "V": 584,
      "T": 106,
      "A": 24,
      "F": 6,
      "F_priority": 3,
      "H_unsupported_claims": 3,
      "points": 1274,
      "max_points": 1440,
      "weighted_percent": 88.47222222222221,
      "fully_correct_percent": 81.11111111111111,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-mincache-2026-10-27-11/summary/results.json",
      "criterion_source": "strata-mincache-2026-10-27-11/summary/fact_reviews.csv",
      "id": "10"
    },
    {
      "corpus": "GP60",
      "cohort": "R12",
      "source_key": "IQ3_XXS",
      "model": "Flash XXS",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS / Strata",
      "profile": "576 MiB Cache",
      "cache_mib": 576,
      "cases": 60,
      "criteria": 720,
      "V": 625,
      "T": 82,
      "A": 5,
      "F": 8,
      "F_priority": 5,
      "H_unsupported_claims": 7,
      "points": 1332,
      "max_points": 1440,
      "weighted_percent": 92.5,
      "fully_correct_percent": 86.80555555555556,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-iq3-2026-10-27-12/summary/comparison.csv",
      "criterion_source": "strata-iq3-2026-10-27-12/summary/fact_reviews.csv",
      "id": "11"
    },
    {
      "corpus": "GP60",
      "cohort": "R13",
      "source_key": "minimum",
      "model": "Flash XXS",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS / Strata",
      "profile": "576 MiB Cache",
      "cache_mib": 576,
      "cases": 60,
      "criteria": 720,
      "V": 625,
      "T": 76,
      "A": 7,
      "F": 12,
      "F_priority": 6,
      "H_unsupported_claims": 17,
      "points": 1326,
      "max_points": 1440,
      "weighted_percent": 92.08333333333333,
      "fully_correct_percent": 86.80555555555556,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-iq3-cache16-2026-10-27-13/summary/results.json",
      "criterion_source": "strata-iq3-cache16-2026-10-27-13/summary/fact_reviews.csv",
      "id": "12"
    },
    {
      "corpus": "GP60",
      "cohort": "R13",
      "source_key": "maximum16",
      "model": "Flash XXS",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ3_XXS / Strata",
      "profile": "1664 MiB Cache",
      "cache_mib": 1664,
      "cases": 60,
      "criteria": 720,
      "V": 622,
      "T": 83,
      "A": 8,
      "F": 7,
      "F_priority": 4,
      "H_unsupported_claims": 24,
      "points": 1327,
      "max_points": 1440,
      "weighted_percent": 92.15277777777777,
      "fully_correct_percent": 86.38888888888889,
      "schema_ok": 60,
      "note": "Grenzprofil: nur 54,2 MiB Reserve bis 16 GiB beim Audiopeak.",
      "same_answers_as": null,
      "summary_source": "strata-iq3-cache16-2026-10-27-13/summary/results.json",
      "criterion_source": "strata-iq3-cache16-2026-10-27-13/summary/fact_reviews.csv",
      "id": "13"
    },
    {
      "corpus": "GP60",
      "cohort": "R12",
      "source_key": "IQ3_S",
      "model": "Flash S",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ3_S / Strata",
      "profile": "640 MiB Cache",
      "cache_mib": 640,
      "cases": 60,
      "criteria": 720,
      "V": 626,
      "T": 80,
      "A": 8,
      "F": 6,
      "F_priority": 4,
      "H_unsupported_claims": 16,
      "points": 1332,
      "max_points": 1440,
      "weighted_percent": 92.5,
      "fully_correct_percent": 86.94444444444444,
      "schema_ok": 60,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-iq3-2026-10-27-12/summary/comparison.csv",
      "criterion_source": "strata-iq3-2026-10-27-12/summary/fact_reviews.csv",
      "id": "14"
    },
    {
      "corpus": "GP60",
      "cohort": "R14",
      "source_key": "iq3s_max24",
      "model": "Flash S",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ3_S / Strata",
      "profile": "9472 MiB Cache",
      "cache_mib": 9472,
      "cases": 60,
      "criteria": 720,
      "V": 612,
      "T": 88,
      "A": 11,
      "F": 9,
      "F_priority": 6,
      "H_unsupported_claims": 14,
      "points": 1312,
      "max_points": 1440,
      "weighted_percent": 91.11111111111111,
      "fully_correct_percent": 85.0,
      "schema_ok": 60,
      "note": "Grenzprofil: nur 36,2 MiB Reserve bis 24 GiB beim Audiopeak.",
      "same_answers_as": null,
      "summary_source": "strata-qwen36-iq3s24-2026-10-28-14/summary/results.json",
      "criterion_source": "strata-qwen36-iq3s24-2026-10-28-14/summary/fact_reviews.csv",
      "id": "15"
    }
  ],
  "gp30_historical": [
    {
      "corpus": "GP30",
      "cohort": "R05",
      "source_key": "4b_q4_8gb",
      "model": "4B-Q4",
      "full_name": "Qwen3.5-4B Q4_K_M",
      "profile": "87552 Token · 1 Slot",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 150,
      "T": 51,
      "A": 36,
      "F": 3,
      "F_priority": null,
      "H_unsupported_claims": 8,
      "points": 351,
      "max_points": 480,
      "weighted_percent": 73.125,
      "fully_correct_percent": 62.5,
      "schema_ok": null,
      "note": "",
      "same_answers_as": null,
      "summary_source": "gp30-2026-10-25-05/summary/model_summary.csv",
      "criterion_source": "gp30-2026-10-25-05/summary/fact_reviews.csv",
      "id": "H01"
    },
    {
      "corpus": "GP30",
      "cohort": "R05",
      "source_key": "9b_q4_8gb",
      "model": "9B-Q4",
      "full_name": "Qwen3.5-9B Q4_K_M",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 185,
      "T": 46,
      "A": 5,
      "F": 4,
      "F_priority": null,
      "H_unsupported_claims": 5,
      "points": 416,
      "max_points": 480,
      "weighted_percent": 86.66666666666667,
      "fully_correct_percent": 77.08333333333334,
      "schema_ok": null,
      "note": "",
      "same_answers_as": null,
      "summary_source": "gp30-2026-10-25-05/summary/model_summary.csv",
      "criterion_source": "gp30-2026-10-25-05/summary/fact_reviews.csv",
      "id": "H02"
    },
    {
      "corpus": "GP30",
      "cohort": "R05",
      "source_key": "9b_q6_original",
      "model": "9B-Q6",
      "full_name": "Qwen3.5-9B Q6_K",
      "profile": "Originalprofil · 3 Slots",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 196,
      "T": 37,
      "A": 4,
      "F": 3,
      "F_priority": null,
      "H_unsupported_claims": 7,
      "points": 429,
      "max_points": 480,
      "weighted_percent": 89.375,
      "fully_correct_percent": 81.66666666666667,
      "schema_ok": null,
      "note": "",
      "same_answers_as": null,
      "summary_source": "gp30-2026-10-25-05/summary/model_summary.csv",
      "criterion_source": "gp30-2026-10-25-05/summary/fact_reviews.csv",
      "id": "H03"
    },
    {
      "corpus": "GP30",
      "cohort": "R05",
      "source_key": "27b_q4_native",
      "model": "27B-Q4",
      "full_name": "Qwen3.8-27B UD-Q4_K_M",
      "profile": "262144 Token · 1 Slot",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 218,
      "T": 20,
      "A": 1,
      "F": 1,
      "F_priority": null,
      "H_unsupported_claims": 8,
      "points": 456,
      "max_points": 480,
      "weighted_percent": 95.0,
      "fully_correct_percent": 90.83333333333333,
      "schema_ok": null,
      "note": "",
      "same_answers_as": null,
      "summary_source": "gp30-2026-10-25-05/summary/model_summary.csv",
      "criterion_source": "gp30-2026-10-25-05/summary/fact_reviews.csv",
      "id": "H04"
    },
    {
      "corpus": "GP30",
      "cohort": "R08",
      "source_key": "strata_qwen36_35b_q4",
      "model": "35B-A3B",
      "full_name": "Qwen3.6-35B-A3B UD-Q4_K_XL / Strata",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 198,
      "T": 36,
      "A": 4,
      "F": 2,
      "F_priority": null,
      "H_unsupported_claims": 7,
      "points": 432,
      "max_points": 480,
      "weighted_percent": 90.0,
      "fully_correct_percent": 82.5,
      "schema_ok": 30,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-2026-10-26-08/summary/model_summary.csv",
      "criterion_source": "strata-2026-10-26-08/summary/fact_reviews.csv",
      "id": "H05"
    },
    {
      "corpus": "GP30",
      "cohort": "R08",
      "source_key": "strata_qwen38_flash_iq2xs",
      "model": "Flash IQ2",
      "full_name": "Qwen3.8-Flash-Next GSQ-RCO IQ2_XS / Strata",
      "profile": "32k · 1 Slot",
      "cache_mib": null,
      "cases": 30,
      "criteria": 240,
      "V": 204,
      "T": 34,
      "A": 2,
      "F": 0,
      "F_priority": null,
      "H_unsupported_claims": 7,
      "points": 442,
      "max_points": 480,
      "weighted_percent": 92.08333333333333,
      "fully_correct_percent": 85.0,
      "schema_ok": 30,
      "note": "",
      "same_answers_as": null,
      "summary_source": "strata-2026-10-26-08/summary/model_summary.csv",
      "criterion_source": "strata-2026-10-26-08/summary/fact_reviews.csv",
      "id": "H06"
    }
  ],
  "gp60_chart_order": {
    "sort": "weighted_percent descending; stable ID for ties",
    "ids": [
      "04",
      "11",
      "14",
      "13",
      "12",
      "05",
      "15",
      "07",
      "08",
      "09",
      "10",
      "06",
      "03",
      "02",
      "01"
    ]
  },
  "source_sha256": {
    "resident-gp60-2026-10-27-09/summary/model_summary.csv": "6f99b1c9f844f4d0b5952245a1b99578eb3fd4252a3c266ba8416f3c3bd08911",
    "resident-gp60-2026-10-27-09/summary/fact_reviews.csv": "5bdc4a93c8115c916da7f8c5daecb5d14d3d00d76c27943ec9d0a9ba8d8c0db2",
    "strata-mincache-2026-10-27-11/summary/results.json": "df8645acb424ac440f470112cd1dfede87f39859d9a198f4987e1b2228850cd0",
    "strata-mincache-2026-10-27-11/summary/fact_reviews.csv": "db40419e59a479136ef43e29752552823db931231a6d1035b5408603659b57da",
    "strata-iq3-2026-10-27-12/summary/comparison.csv": "84d9289e77a2c705dcdac954d9e01ce6b6c7fa30e09e67c9dc02894b932eba01",
    "strata-iq3-2026-10-27-12/summary/fact_reviews.csv": "ee6a9f6aa64c02e295668deecef9f42a004f100040a1598badc203fff5c49e0f",
    "strata-iq3-cache16-2026-10-27-13/summary/results.json": "5f6f8380289b44d55e84d0ed57d322d663ebbdd51cef02db8417debffd258aa9",
    "strata-iq3-cache16-2026-10-27-13/summary/fact_reviews.csv": "8cdedba8ee1171818473d365e23265089c2a05bc19a26da61055eba6ac138365",
    "strata-qwen36-iq3s24-2026-10-28-14/summary/results.json": "811d16aabb4a99bcd1d88e07b1e8a26e8149f6629fb8153bd3ce75183affd252",
    "strata-qwen36-iq3s24-2026-10-28-14/summary/fact_reviews.csv": "e5ae90c8432268112632300c0cac5bf15ad64b209752339def807fd4512b2f67",
    "gp30-2026-10-25-05/summary/model_summary.csv": "a0d92e294a8cf567f3157edb580f6e198d2cd7155e4a258f71a74718998cd92d",
    "gp30-2026-10-25-05/summary/fact_reviews.csv": "301613f12673674fcf11fe47e240535a1d89991d388bb77b7a0b1c70931e5d4f",
    "strata-2026-10-26-08/summary/model_summary.csv": "1f9ca2086a5760894e2df51a1a80627a56bea421ccd19f3a8a2f731835b38f92",
    "strata-2026-10-26-08/summary/fact_reviews.csv": "3786c3a325aa12eaa857d4e22a48e84b4095ad84aad32ede13af2ff0982d0ed3",
    "hardware-matrix-2026-10-28/Hardwarematrix_mit_Zeiten.html": "78b70c91c728c153cc5d5b8155763d9d560dda615885353b41cf197c0d5dcfe3",
    "resident-gp60-2026-10-27-09/corpus/cases.json": "4cf9c39162037777875a76e8b943a09deb2cdc06383c14efdd5d03eedc48bfd8"
  },
  "validation": {
    "primary_aggregate_matches_all_individual_grades": true,
    "priority_F_matches_individual_grades_where_available": true,
    "gp60_review_rows": 15,
    "gp30_review_rows": 6,
    "criterion_judgments_checked": 12240,
    "same_answers_reviewed_twice": [
      "R09/27b_q4",
      "R12/27b_q4"
    ],
    "copied_comparison_rows_excluded": true
  }
}