{
  "schemaVersion": 1,
  "asOfDate": "2026-09-25",
  "dimensions": [
    "readability",
    "warmth",
    "directness",
    "absenceOfAiIsms",
    "toneAdherence",
    "factualRestraint"
  ],
  "weights": {
    "readability": 1,
    "warmth": 1,
    "directness": 1,
    "absenceOfAiIsms": 1,
    "toneAdherence": 1,
    "factualRestraint": 1
  },
  "records": [
    {
      "id": "seed-gpt-6-astra-de",
      "modelId": "gpt-6-astra",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-6-astra-en",
      "modelId": "gpt-6-astra",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-gpt-6-sol-de",
      "modelId": "gpt-6-sol",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-6-sol-en",
      "modelId": "gpt-6-sol",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-gpt-6-luna-de",
      "modelId": "gpt-6-luna",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-6-luna-en",
      "modelId": "gpt-6-luna",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-sol-de",
      "modelId": "gpt-5.6-sol",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-sol-en",
      "modelId": "gpt-5.6-sol",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-terra-de",
      "modelId": "gpt-5.6-terra",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-terra-en",
      "modelId": "gpt-5.6-terra",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-luna-de",
      "modelId": "gpt-5.6-luna",
      "cliId": "codex",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-gpt-5.6-luna-en",
      "modelId": "gpt-5.6-luna",
      "cliId": "codex",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-claude-opus-5-5-de",
      "modelId": "claude-opus-5-5",
      "cliId": "claude-code",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-claude-opus-5-5-en",
      "modelId": "claude-opus-5-5",
      "cliId": "claude-code",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-claude-opus-5-de",
      "modelId": "claude-opus-5",
      "cliId": "claude-code",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-claude-opus-5-en",
      "modelId": "claude-opus-5",
      "cliId": "claude-code",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-claude-sonnet-5-de",
      "modelId": "claude-sonnet-5",
      "cliId": "claude-code",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-claude-sonnet-5-en",
      "modelId": "claude-sonnet-5",
      "cliId": "claude-code",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-claude-haiku-4-5-de",
      "modelId": "claude-haiku-4-5",
      "cliId": "claude-code",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-claude-haiku-4-5-en",
      "modelId": "claude-haiku-4-5",
      "cliId": "claude-code",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    },
    {
      "id": "seed-claude-fable-5-1-de",
      "modelId": "claude-fable-5-1",
      "cliId": "claude-code",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.ltedi-1.9/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification."
        }
      ]
    },
    {
      "id": "seed-claude-fable-5-1-en",
      "modelId": "claude-fable-5-1",
      "cliId": "claude-code",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "scores": {
        "readability": null,
        "warmth": null,
        "directness": null,
        "absenceOfAiIsms": null,
        "toneAdherence": null,
        "factualRestraint": null
      },
      "overall": null,
      "costPerSampleUsd": null,
      "notes": "Research-informed evidence gap: these studies inform the rubric, but do not measure this exact model/language/thinking-level combination. Await Voice Lint; no numeric extrapolation.",
      "evidence": [
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.emnlp-main.519/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Use multiple dimensions and human spot checks; not a score for any seeded model."
        },
        {
          "kind": "study",
          "reference": "https://aclanthology.org/2024.tacl-1.24/",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Assess readability and factual restraint separately; no exact-model transfer."
        },
        {
          "kind": "study",
          "reference": "https://arxiv.org/abs/2404.04475",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric."
        },
        {
          "kind": "study",
          "reference": "https://www.nature.com/articles/s41586-026-10410-0",
          "observedAtUtc": "2026-09-25T00:00:00Z",
          "note": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models."
        }
      ]
    }
  ],
  "studies": [
    {
      "id": "appls-2024",
      "title": "APPLS: Evaluating Evaluation Metrics for Plain Language Summarization",
      "authors": "Yue Guo; Tal August; Gondy Leroy; Trevor Cohen; Lucy Lu Wang",
      "year": 2024,
      "url": "https://aclanthology.org/2024.emnlp-main.519/",
      "claim": "No single tested metric captured all four plain-language criteria.",
      "relevance": "Use multiple dimensions and human spot checks; not a score for any seeded model.",
      "kind": "peer-reviewed",
      "status": "provisional",
      "verifiedAtUtc": "2026-09-25T00:00:00Z"
    },
    {
      "id": "meaning-2024",
      "title": "Do Text Simplification Systems Preserve Meaning? A Human Evaluation via Reading Comprehension",
      "authors": "Sweta Agrawal; Marine Carpuat",
      "year": 2024,
      "url": "https://aclanthology.org/2024.tacl-1.24/",
      "claim": "Simpler text can omit information needed to answer comprehension questions.",
      "relevance": "Assess readability and factual restraint separately; no exact-model transfer.",
      "kind": "peer-reviewed",
      "status": "provisional",
      "verifiedAtUtc": "2026-09-25T00:00:00Z"
    },
    {
      "id": "length-2024",
      "title": "Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators",
      "authors": "Yann Dubois; Bal\u00e1zs Galambosi; Percy Liang; Tatsunori B. Hashimoto",
      "year": 2024,
      "url": "https://arxiv.org/abs/2404.04475",
      "claim": "Controlling response length reduces verbosity bias in automatic preference judging.",
      "relevance": "Balance lengths and randomize blinded comparisons; preference scores are not our six-dimension rubric.",
      "kind": "peer-reviewed",
      "status": "provisional",
      "verifiedAtUtc": "2026-09-25T00:00:00Z"
    },
    {
      "id": "warmth-2026",
      "title": "Training language models to be warm can reduce accuracy and increase sycophancy",
      "authors": "Lujain Ibrahim; Franziska Sofia Hafner; Luc Rocher",
      "year": 2026,
      "url": "https://www.nature.com/articles/s41586-026-10410-0",
      "claim": "Warmth fine-tuning reduced accuracy and increased sycophancy in the tested models.",
      "relevance": "Keep a factual-restraint floor alongside warmth; these fine-tuned models are not the catalogue models.",
      "kind": "peer-reviewed",
      "status": "provisional",
      "verifiedAtUtc": "2026-09-25T00:00:00Z"
    },
    {
      "id": "de-lite-2024",
      "title": "DE-Lite - a New Corpus of Easy German: Compilation, Exploration, Analysis",
      "authors": "Sarah Jablotschkin; Elke Teich; Heike Zinsmeister",
      "year": 2024,
      "url": "https://aclanthology.org/2024.ltedi-1.9/",
      "claim": "Genre affects the linguistic differences between Easy and Standard German.",
      "relevance": "Evaluate German separately across text genres; plain friendly language is not an Easy German certification.",
      "kind": "peer-reviewed",
      "status": "provisional",
      "verifiedAtUtc": "2026-09-25T00:00:00Z"
    }
  ],
  "source": "src/TokenEconomy/catalog/language-capabilities.json",
  "selectedRecordIds": [
    "seed-claude-fable-5-1-de",
    "seed-claude-fable-5-1-en",
    "seed-claude-haiku-4-5-de",
    "seed-claude-haiku-4-5-en",
    "seed-claude-opus-5-5-de",
    "seed-claude-opus-5-5-en",
    "seed-claude-opus-5-de",
    "seed-claude-opus-5-en",
    "seed-claude-sonnet-5-de",
    "seed-claude-sonnet-5-en",
    "seed-gpt-5.6-luna-de",
    "seed-gpt-5.6-luna-en",
    "seed-gpt-5.6-sol-de",
    "seed-gpt-5.6-sol-en",
    "seed-gpt-5.6-terra-de",
    "seed-gpt-5.6-terra-en",
    "seed-gpt-6-astra-de",
    "seed-gpt-6-astra-en",
    "seed-gpt-6-luna-de",
    "seed-gpt-6-luna-en",
    "seed-gpt-6-sol-de",
    "seed-gpt-6-sol-en"
  ],
  "languagePriors": [
    {
      "textKind": "copy",
      "taskClass": "docEdit",
      "language": "de",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "de",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "copy",
      "taskClass": "docEdit",
      "language": "en",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "en",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "documentation",
      "taskClass": "docEdit",
      "language": "de",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "de",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "documentation",
      "taskClass": "docEdit",
      "language": "en",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "en",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "operatorMessages",
      "taskClass": "docEdit",
      "language": "de",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "de",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "operatorMessages",
      "taskClass": "docEdit",
      "language": "en",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "en",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "replies",
      "taskClass": "docEdit",
      "language": "de",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "de",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    },
    {
      "textKind": "replies",
      "taskClass": "docEdit",
      "language": "en",
      "asOfDate": "2026-09-25",
      "requirement": {
        "language": "en",
        "minimumOverall": 0.75,
        "minimumScores": {
          "readability": 0.75,
          "absenceOfAiIsms": 0.75,
          "factualRestraint": 0.8
        },
        "thinkingLevel": "medium",
        "allowProvisional": false
      },
      "candidates": [],
      "note": "No retained measured row meets the requirement yet. Import Voice Lint evidence; an empty set means wait, not an invented cheapest model."
    }
  ],
  "costQuality": [
    {
      "recordId": "seed-claude-fable-5-1-de",
      "modelId": "claude-fable-5-1",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-fable-5-1-en",
      "modelId": "claude-fable-5-1",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-haiku-4-5-de",
      "modelId": "claude-haiku-4-5",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-haiku-4-5-en",
      "modelId": "claude-haiku-4-5",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-opus-5-5-de",
      "modelId": "claude-opus-5-5",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-opus-5-5-en",
      "modelId": "claude-opus-5-5",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-opus-5-de",
      "modelId": "claude-opus-5",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-opus-5-en",
      "modelId": "claude-opus-5",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-sonnet-5-de",
      "modelId": "claude-sonnet-5",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-claude-sonnet-5-en",
      "modelId": "claude-sonnet-5",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-luna-de",
      "modelId": "gpt-5.6-luna",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-luna-en",
      "modelId": "gpt-5.6-luna",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-sol-de",
      "modelId": "gpt-5.6-sol",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-sol-en",
      "modelId": "gpt-5.6-sol",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-terra-de",
      "modelId": "gpt-5.6-terra",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-5.6-terra-en",
      "modelId": "gpt-5.6-terra",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-astra-de",
      "modelId": "gpt-6-astra",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-astra-en",
      "modelId": "gpt-6-astra",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-luna-de",
      "modelId": "gpt-6-luna",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-luna-en",
      "modelId": "gpt-6-luna",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-sol-de",
      "modelId": "gpt-6-sol",
      "language": "de",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    },
    {
      "recordId": "seed-gpt-6-sol-en",
      "modelId": "gpt-6-sol",
      "language": "en",
      "thinkingLevel": "unspecified",
      "status": "provisional",
      "overall": null,
      "costPerSampleUsd": null,
      "costPerQualityPointUsd": null
    }
  ]
}