{
  "release": {
    "id": "CV-LRB-2026.08.15-R13",
    "date": "2026-08-15",
    "archive_filename": "CloseVector-Legal-Reasoning-Benchmark-CV-LRB-2026.08.15-R13.zip",
    "supersedes": "CV-LRB-2026.08.15-R12"
  },
  "headline": {
    "leader": "Qwen 3.8 27B",
    "rounded_equal_or_better_pct": 98,
    "exact_equal_or_better_pct": 98.33333333333333,
    "equal_or_better": 236,
    "total_judgments_involving_leader": 240,
    "wins": 30,
    "losses": 4,
    "ties": 206,
    "label": "matched or beat all three opponents"
  },
  "record_counts": {
    "questions": 80,
    "models": 4,
    "answers": 320,
    "judge_reports": 480,
    "reasoning_traces": 309,
    "pairings_per_matter": 6
  },
  "model_standings": {
    "qwen38": {
      "display": "Qwen 3.8 27B",
      "wins": 30,
      "losses": 4,
      "ties": 206,
      "both_materially_inadequate": 0,
      "equal_or_better": 236,
      "total_judgments": 240,
      "equal_or_better_pct": 98.33333333333333,
      "rejected_points": 158
    },
    "deepseek": {
      "display": "DeepSeek V4-Flash-0731",
      "wins": 23,
      "losses": 8,
      "ties": 207,
      "both_materially_inadequate": 2,
      "equal_or_better": 230,
      "total_judgments": 240,
      "equal_or_better_pct": 95.83333333333334,
      "rejected_points": 173
    },
    "claude": {
      "display": "Claude Opus 5",
      "wins": 23,
      "losses": 18,
      "ties": 199,
      "both_materially_inadequate": 0,
      "equal_or_better": 222,
      "total_judgments": 240,
      "equal_or_better_pct": 92.5,
      "rejected_points": 565
    },
    "qwen35": {
      "display": "Qwen 3.5 27B",
      "wins": 4,
      "losses": 50,
      "ties": 184,
      "both_materially_inadequate": 2,
      "equal_or_better": 188,
      "total_judgments": 240,
      "equal_or_better_pct": 78.33333333333333,
      "rejected_points": 507
    }
  },
  "head_to_head": {
    "qwen38-claude": {
      "first_model": "Qwen 3.8 27B",
      "second_model": "Claude Opus 5",
      "first_wins": 6,
      "second_wins": 4,
      "ties": 70,
      "both_materially_inadequate": 0,
      "first_equal_or_better": 76,
      "total": 80,
      "first_equal_or_better_pct": 95.0
    },
    "qwen38-deepseek": {
      "first_model": "Qwen 3.8 27B",
      "second_model": "DeepSeek V4-Flash-0731",
      "first_wins": 5,
      "second_wins": 0,
      "ties": 75,
      "both_materially_inadequate": 0,
      "first_equal_or_better": 80,
      "total": 80,
      "first_equal_or_better_pct": 100.0
    },
    "qwen38-qwen35": {
      "first_model": "Qwen 3.8 27B",
      "second_model": "Qwen 3.5 27B",
      "first_wins": 19,
      "second_wins": 0,
      "ties": 61,
      "both_materially_inadequate": 0,
      "first_equal_or_better": 80,
      "total": 80,
      "first_equal_or_better_pct": 100.0
    },
    "deepseek-claude": {
      "first_model": "DeepSeek V4-Flash-0731",
      "second_model": "Claude Opus 5",
      "first_wins": 8,
      "second_wins": 3,
      "ties": 69,
      "both_materially_inadequate": 0,
      "first_equal_or_better": 77,
      "total": 80,
      "first_equal_or_better_pct": 96.25
    },
    "deepseek-qwen35": {
      "first_model": "DeepSeek V4-Flash-0731",
      "second_model": "Qwen 3.5 27B",
      "first_wins": 15,
      "second_wins": 0,
      "ties": 63,
      "both_materially_inadequate": 2,
      "first_equal_or_better": 78,
      "total": 80,
      "first_equal_or_better_pct": 97.5
    },
    "qwen35-claude": {
      "first_model": "Qwen 3.5 27B",
      "second_model": "Claude Opus 5",
      "first_wins": 4,
      "second_wins": 16,
      "ties": 60,
      "both_materially_inadequate": 0,
      "first_equal_or_better": 64,
      "total": 80,
      "first_equal_or_better_pct": 80.0
    }
  },
  "series": {
    "series_1": {
      "wins": 15,
      "losses": 1,
      "ties": 44,
      "equal_or_better": 59,
      "total": 60,
      "pct": 98.33333333333333
    },
    "series_2": {
      "wins": 15,
      "losses": 3,
      "ties": 162,
      "equal_or_better": 177,
      "total": 180,
      "pct": 98.33333333333333
    }
  },
  "qwen38_vs_claude_completed_output_check": {
    "excluded_matters": [
      "q59",
      "q61",
      "q75"
    ],
    "wins": 4,
    "losses": 4,
    "ties": 69,
    "total": 77
  },
  "judge_prompt": {
    "page": "judge-prompt/index.html",
    "sha256": "7e58b951cb86e4deba3b60fd01e9b9bd247ebe9883e1c9e43b24acffb1fd4667",
    "published_verdict_records": 480
  },
  "delivery_record": {
    "qwen38": {
      "stop": 80,
      "failed": 0,
      "truncated": 0
    },
    "hosted_claude_delivery_path": {
      "content_filter_matters": [
        "q59",
        "q61",
        "q75"
      ]
    }
  },
  "result_lineage": {
    "original_headline": {
      "comparison": "DeepSeek V4-Flash-0731 vs Claude Opus 5",
      "equal_or_better": 77,
      "total_matters": 80,
      "displayed_pct": 96,
      "unit": "matters"
    },
    "expanded_headline": {
      "leader": "Qwen 3.8 27B",
      "equal_or_better": 236,
      "total_judgments": 240,
      "displayed_pct": 98,
      "unit": "judgments across three opponents"
    }
  },
  "qwen38_run_conditions": {
    "model_file": "Qwen3.8-27B-UD-Q6_K_XL.gguf",
    "artifact_quantization_label": "UD-Q6_K_XL",
    "server_reported_model_ftype": "Q4_K - Small",
    "execution": "one local GPU",
    "reasoning_effort": "xhigh",
    "temperature": 1.0,
    "top_k": 20,
    "top_p": 0.949999988079071,
    "presence_penalty": 0.0,
    "chat_template_sha256_prefix": "12827f24b742ea4e",
    "completion_tokens": 2206191,
    "completed_runs": 80,
    "truncated_runs": 0,
    "generation_started": "2026-08-14T12:53:36-0400",
    "generation_finished": "2026-08-15T00:24:33-0400",
    "new_judgments_completed_date": "2026-08-15"
  },
  "publication_timing": {
    "complete_80_matter_release_date": "2026-08-10",
    "qwen38_generation_started": "2026-08-14T12:53:36-0400",
    "note": "The first 20 matters were public before the complete 80-matter release."
  }
}