{
  "rodada": "2026-09-24",
  "executado_por": "LUA Vision",
  "rotulo": "Registro LUA · não é avaliação independente",
  "ic95": "bootstrap percentil, 4000 reamostragens, semente crc32('bench:coluna'); ticket por modelo de pergunta",
  "benchmarks": {
    "enamed": {
      "n": 87,
      "colunas": {
        "house": {
          "rotulo": "LUA Genesys PI House",
          "score": 96.6,
          "ic95": [
            92.0,
            100.0
          ],
          "respondidas": 85,
          "recusadas_ou_barradas": 2,
          "sem_resposta": 0
        },
        "gpt54": {
          "rotulo": "GPT-5.4 (API comercial)",
          "score": 94.3,
          "ic95": [
            88.5,
            98.9
          ],
          "respondidas": 85,
          "recusadas_ou_barradas": 2,
          "sem_resposta": 0
        },
        "gpt55": {
          "rotulo": "GPT-5.5 (API comercial)",
          "score": 96.6,
          "ic95": [
            92.0,
            100.0
          ],
          "respondidas": 85,
          "recusadas_ou_barradas": 2,
          "sem_resposta": 0
        },
        "grok": {
          "rotulo": "Grok 4.6 (API comercial)",
          "score": 95.4,
          "ic95": [
            90.8,
            98.9
          ],
          "respondidas": 87,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "deepseek": {
          "rotulo": "DeepSeek-V4-Pro (API comercial)",
          "score": 94.3,
          "ic95": [
            88.5,
            98.9
          ],
          "respondidas": 87,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "kimi": {
          "rotulo": "Kimi K2.6 (API comercial)",
          "score": 93.1,
          "ic95": [
            87.4,
            97.7
          ],
          "respondidas": 86,
          "recusadas_ou_barradas": 1,
          "sem_resposta": 0
        }
      }
    },
    "poscomp": {
      "n": 130,
      "colunas": {
        "house": {
          "rotulo": "LUA Genesys PI House",
          "score": 95.4,
          "ic95": [
            91.5,
            98.5
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "gpt54": {
          "rotulo": "GPT-5.4 (API comercial)",
          "score": 95.4,
          "ic95": [
            91.5,
            98.5
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "gpt55": {
          "rotulo": "GPT-5.5 (API comercial)",
          "score": 95.4,
          "ic95": [
            91.5,
            98.5
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "grok": {
          "rotulo": "Grok 4.6 (API comercial)",
          "score": 96.2,
          "ic95": [
            92.3,
            99.2
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "deepseek": {
          "rotulo": "DeepSeek-V4-Pro (API comercial)",
          "score": 93.1,
          "ic95": [
            88.5,
            96.9
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "kimi": {
          "rotulo": "Kimi K2.6 (API comercial)",
          "score": 95.4,
          "ic95": [
            91.5,
            98.5
          ],
          "respondidas": 130,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        }
      }
    },
    "bluex": {
      "n": 809,
      "colunas": {
        "house": {
          "rotulo": "LUA Genesys PI House",
          "score": 96.0,
          "ic95": [
            94.6,
            97.3
          ],
          "respondidas": 802,
          "recusadas_ou_barradas": 7,
          "sem_resposta": 0
        },
        "gpt54": {
          "rotulo": "GPT-5.4 (API comercial)",
          "score": 95.4,
          "ic95": [
            93.9,
            96.8
          ],
          "respondidas": 802,
          "recusadas_ou_barradas": 7,
          "sem_resposta": 0
        },
        "gpt55": {
          "rotulo": "GPT-5.5 (API comercial)",
          "score": 96.3,
          "ic95": [
            94.9,
            97.5
          ],
          "respondidas": 802,
          "recusadas_ou_barradas": 7,
          "sem_resposta": 0
        },
        "grok": {
          "rotulo": "Grok 4.6 (API comercial)",
          "score": 96.3,
          "ic95": [
            94.9,
            97.5
          ],
          "respondidas": 805,
          "recusadas_ou_barradas": 1,
          "sem_resposta": 3
        },
        "deepseek": {
          "rotulo": "DeepSeek-V4-Pro (API comercial)",
          "score": 93.8,
          "ic95": [
            92.1,
            95.4
          ],
          "respondidas": 807,
          "recusadas_ou_barradas": 2,
          "sem_resposta": 0
        },
        "kimi": {
          "rotulo": "Kimi K2.6 (API comercial)",
          "score": 94.1,
          "ic95": [
            92.5,
            95.6
          ],
          "respondidas": 797,
          "recusadas_ou_barradas": 12,
          "sem_resposta": 0
        }
      }
    },
    "ticket": {
      "n": 151,
      "colunas": {
        "house": {
          "rotulo": "LUA Genesys PI House",
          "score": 87.4,
          "ic95": [
            76.0,
            96.8
          ],
          "respondidas": 151,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "gpt54": {
          "rotulo": "GPT-5.4 (API comercial)",
          "score": 86.8,
          "ic95": [
            74.8,
            96.6
          ],
          "respondidas": 151,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "gpt55": {
          "rotulo": "GPT-5.5 (API comercial)",
          "score": 86.1,
          "ic95": [
            73.9,
            96.5
          ],
          "respondidas": 151,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "grok": {
          "rotulo": "Grok 4.6 (API comercial)",
          "score": null,
          "ic95": null,
          "respondidas": 39,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 112,
          "fora_da_tabela": "Erro do ambiente com page_number: toda chamada falhou. Ver defeitos."
        },
        "deepseek": {
          "rotulo": "DeepSeek-V4-Pro (API comercial)",
          "score": 85.4,
          "ic95": [
            74.5,
            95.1
          ],
          "respondidas": 151,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        },
        "kimi": {
          "rotulo": "Kimi K2.6 (API comercial)",
          "score": 84.8,
          "ic95": [
            73.0,
            95.3
          ],
          "respondidas": 151,
          "recusadas_ou_barradas": 0,
          "sem_resposta": 0
        }
      }
    }
  }
}
