{
  "name": "The Latent Score",
  "release": "tli-2026-v1.0-2026-09-10-benchmarks-1",
  "methodologyVersion": "tli-2026-v1.0",
  "dateModified": "2026-09-10",
  "publisher": "The Latent",
  "source": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1",
  "currentRankings": "https://thelatent.co/score",
  "schemaVersion": 2,
  "scale": {
    "mean": 100,
    "sd": 15,
    "unit": "points",
    "description": "Frozen calibration-cohort reference scale; not human IQ or a percentage.",
    "cohortSize": 18
  },
  "definitions": {
    "rank": "Models are ranked by the displayed point score once they have four contributing benchmarks across two domains. Ties use stable model-name ordering. Ranks are a current ordering, not a confidence test. Scores are published immediately without a provisional label; early scores can move substantially as evidence arrives. Contributing benchmark counts and measured domain coverage appear beside each score.",
    "stabilityRange": "The score stability range is a 95% bootstrap score interval conditional on frozen calibration and comparable evidence availability. It resamples benchmark families and evaluators and adds a frozen missing-evidence correction. It excludes selection across reasoning settings, selective publication, calibration changes and new domains. It is not a 95% guarantee about true capability or where a future score will land. Sparse or conflicting evidence can widen the range.",
    "domainScore": "Standard-deviation units within a domain relative to the calibration cohort. Null means no measured score; it is not zero.",
    "benchmarksUsed": "Number of results contributing to the score, not a count of independent experiments.",
    "reportingLimitations": "The separate reporting sensitivity range tests what happens when only the strongest or weakest standardized benchmark results are retained. It has no probability interpretation and is not combined with the score stability range. Stress-test coverage describes recovery of fuller-data scores under these deliberate masks, not real-world coverage or true intelligence. This test does not quantify the separate advantage from trying more reasoning settings."
  },
  "domains": [
    {
      "id": "knowledge_reasoning",
      "name": "Knowledge & reasoning",
      "description": "Academic knowledge, mathematics, abstract reasoning, factuality, and long-context reasoning.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 17
    },
    {
      "id": "coding_swe",
      "name": "Coding & software engineering",
      "description": "Agentic terminal work, issue resolution, long-horizon engineering, competitive coding, and autonomous ML engineering.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 19
    },
    {
      "id": "agentic_tooluse",
      "name": "Agentic tool & computer use",
      "description": "Tool orchestration, MCP servers, business-workflow automation, browsing, and GUI computer use.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 16
    },
    {
      "id": "professional_work",
      "name": "Professional & real-world work",
      "description": "Economically valuable knowledge work: legal, finance, tax, medical, enterprise documents, and occupational workflows.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 21
    },
    {
      "id": "multimodal",
      "name": "Multimodal & vision",
      "description": "Chart, image, video, and handwriting understanding.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 10
    },
    {
      "id": "security",
      "name": "Cybersecurity",
      "description": "Vulnerability analysis and exploit development capability.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 2
    },
    {
      "id": "preference_communication",
      "name": "Preference & communication",
      "description": "Arena-style human preference, emotional intelligence, creative writing, and design preference.",
      "unit": "calibration-cohort standard deviations",
      "benchmarkCount": 4
    }
  ],
  "methodology": "https://thelatent.co/methodology",
  "license": "https://thelatent.co/terms-service#data-usage",
  "models": [
    {
      "modelId": "gpt-6-astra",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gpt-6-astra",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gpt-6-astra",
      "name": "GPT-6 Astra",
      "provider": "OpenAI",
      "rank": 1,
      "score": 140.8,
      "stabilityRange": [
        130.1,
        151.5
      ],
      "provisional": false,
      "benchmarksUsed": 53,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 1.4668,
          "benchmarksObserved": 16
        },
        {
          "domainId": "coding_swe",
          "score": 1.2316,
          "benchmarksObserved": 11
        },
        {
          "domainId": "agentic_tooluse",
          "score": 1.3026,
          "benchmarksObserved": 7
        },
        {
          "domainId": "professional_work",
          "score": 0.6106,
          "benchmarksObserved": 13
        },
        {
          "domainId": "multimodal",
          "score": 1.1028,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": 1.0121,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": 1.0551,
          "benchmarksObserved": 1
        }
      ]
    },
    {
      "modelId": "fable-5-1",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/fable-5-1",
      "currentEvidenceUrl": "https://thelatent.co/score/models/fable-5-1",
      "name": "Fable 5.1",
      "provider": "Anthropic",
      "rank": 2,
      "score": 134.6,
      "stabilityRange": [
        125,
        143.6
      ],
      "provisional": false,
      "benchmarksUsed": 39,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 1.1312,
          "benchmarksObserved": 10
        },
        {
          "domainId": "coding_swe",
          "score": 1.1614,
          "benchmarksObserved": 11
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.91,
          "benchmarksObserved": 3
        },
        {
          "domainId": "professional_work",
          "score": 0.912,
          "benchmarksObserved": 10
        },
        {
          "domainId": "multimodal",
          "score": 0.6663,
          "benchmarksObserved": 3
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": 0.9897,
          "benchmarksObserved": 2
        }
      ]
    },
    {
      "modelId": "opus-5",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/opus-5",
      "currentEvidenceUrl": "https://thelatent.co/score/models/opus-5",
      "name": "Opus 5",
      "provider": "Anthropic",
      "rank": 3,
      "score": 129.5,
      "stabilityRange": [
        123.8,
        135.1
      ],
      "provisional": false,
      "benchmarksUsed": 68,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.9325,
          "benchmarksObserved": 16
        },
        {
          "domainId": "coding_swe",
          "score": 0.8837,
          "benchmarksObserved": 16
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.7951,
          "benchmarksObserved": 11
        },
        {
          "domainId": "professional_work",
          "score": 0.8316,
          "benchmarksObserved": 16
        },
        {
          "domainId": "multimodal",
          "score": 0.7427,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": 0.6189,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": 0.8034,
          "benchmarksObserved": 4
        }
      ]
    },
    {
      "modelId": "fable-5",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/fable-5",
      "currentEvidenceUrl": "https://thelatent.co/score/models/fable-5",
      "name": "Fable 5",
      "provider": "Anthropic",
      "rank": 4,
      "score": 119.6,
      "stabilityRange": [
        112.5,
        126.6
      ],
      "provisional": false,
      "benchmarksUsed": 56,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.6715,
          "benchmarksObserved": 12
        },
        {
          "domainId": "coding_swe",
          "score": 0.6501,
          "benchmarksObserved": 14
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.1512,
          "benchmarksObserved": 8
        },
        {
          "domainId": "professional_work",
          "score": 0.6767,
          "benchmarksObserved": 15
        },
        {
          "domainId": "multimodal",
          "score": 0.5423,
          "benchmarksObserved": 3
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": 0.5601,
          "benchmarksObserved": 4
        }
      ]
    },
    {
      "modelId": "muse-spark-1-3",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/muse-spark-1-3",
      "currentEvidenceUrl": "https://thelatent.co/score/models/muse-spark-1-3",
      "name": "Muse Spark 1.3",
      "provider": "Meta",
      "rank": 5,
      "score": 114.2,
      "stabilityRange": [
        102.3,
        125.8
      ],
      "provisional": false,
      "benchmarksUsed": 13,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.5753,
          "benchmarksObserved": 4
        },
        {
          "domainId": "coding_swe",
          "score": 0.4585,
          "benchmarksObserved": 3
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.3387,
          "benchmarksObserved": 2
        },
        {
          "domainId": "professional_work",
          "score": 0.525,
          "benchmarksObserved": 2
        },
        {
          "domainId": "multimodal",
          "score": 0.1176,
          "benchmarksObserved": 2
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": null,
          "benchmarksObserved": 0
        }
      ]
    },
    {
      "modelId": "gpt-5-6-sol",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gpt-5-6-sol",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gpt-5-6-sol",
      "name": "GPT-5.6 Sol",
      "provider": "OpenAI",
      "rank": 6,
      "score": 111.3,
      "stabilityRange": [
        103.2,
        119.2
      ],
      "provisional": false,
      "benchmarksUsed": 86,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.5497,
          "benchmarksObserved": 16
        },
        {
          "domainId": "coding_swe",
          "score": 0.1841,
          "benchmarksObserved": 19
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.3473,
          "benchmarksObserved": 15
        },
        {
          "domainId": "professional_work",
          "score": 0.1752,
          "benchmarksObserved": 21
        },
        {
          "domainId": "multimodal",
          "score": 0.4361,
          "benchmarksObserved": 10
        },
        {
          "domainId": "security",
          "score": 0.2911,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": 0.1438,
          "benchmarksObserved": 4
        }
      ]
    },
    {
      "modelId": "grok-4-6-high",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/grok-4-6-high",
      "currentEvidenceUrl": "https://thelatent.co/score/models/grok-4-6-high",
      "name": "Grok 4.6",
      "provider": "xAI",
      "rank": 7,
      "score": 105.7,
      "stabilityRange": [
        97.1,
        115.5
      ],
      "provisional": false,
      "benchmarksUsed": 27,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.1745,
          "benchmarksObserved": 6
        },
        {
          "domainId": "coding_swe",
          "score": 0.3539,
          "benchmarksObserved": 9
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.3933,
          "benchmarksObserved": 4
        },
        {
          "domainId": "professional_work",
          "score": -0.0143,
          "benchmarksObserved": 4
        },
        {
          "domainId": "multimodal",
          "score": 0.0876,
          "benchmarksObserved": 2
        },
        {
          "domainId": "security",
          "score": 0.0829,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": -0.0261,
          "benchmarksObserved": 1
        }
      ]
    },
    {
      "modelId": "kimi-k3",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/kimi-k3",
      "currentEvidenceUrl": "https://thelatent.co/score/models/kimi-k3",
      "name": "Kimi K3",
      "provider": "Moonshot AI",
      "rank": 8,
      "score": 104.9,
      "stabilityRange": [
        95.4,
        113.7
      ],
      "provisional": false,
      "benchmarksUsed": 64,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.0956,
          "benchmarksObserved": 14
        },
        {
          "domainId": "coding_swe",
          "score": 0.1516,
          "benchmarksObserved": 11
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.02,
          "benchmarksObserved": 11
        },
        {
          "domainId": "professional_work",
          "score": 0.0289,
          "benchmarksObserved": 16
        },
        {
          "domainId": "multimodal",
          "score": 0.2011,
          "benchmarksObserved": 8
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": 0.5201,
          "benchmarksObserved": 4
        }
      ]
    },
    {
      "modelId": "gemini-3-8-flash",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gemini-3-8-flash",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gemini-3-8-flash",
      "name": "Gemini 3.8 Flash",
      "provider": "Google",
      "rank": 9,
      "score": 102.8,
      "stabilityRange": [
        90.4,
        113.4
      ],
      "provisional": false,
      "benchmarksUsed": 40,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.3258,
          "benchmarksObserved": 6
        },
        {
          "domainId": "coding_swe",
          "score": 0.0427,
          "benchmarksObserved": 12
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.2563,
          "benchmarksObserved": 4
        },
        {
          "domainId": "professional_work",
          "score": -0.2027,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": -0.045,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": null,
          "benchmarksObserved": 0
        }
      ]
    },
    {
      "modelId": "glm-5-3",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/glm-5-3",
      "currentEvidenceUrl": "https://thelatent.co/score/models/glm-5-3",
      "name": "GLM-5.3",
      "provider": "Z.ai",
      "rank": 10,
      "score": 101.6,
      "stabilityRange": [
        90.8,
        113.9
      ],
      "provisional": false,
      "benchmarksUsed": 39,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.1612,
          "benchmarksObserved": 8
        },
        {
          "domainId": "coding_swe",
          "score": 0.169,
          "benchmarksObserved": 11
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.0717,
          "benchmarksObserved": 4
        },
        {
          "domainId": "professional_work",
          "score": 0.1416,
          "benchmarksObserved": 12
        },
        {
          "domainId": "multimodal",
          "score": -0.0877,
          "benchmarksObserved": 1
        },
        {
          "domainId": "security",
          "score": -0.0892,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": 0.2367,
          "benchmarksObserved": 2
        }
      ]
    },
    {
      "modelId": "gemini-3-7-flash",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gemini-3-7-flash",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gemini-3-7-flash",
      "name": "Gemini 3.7 Flash",
      "provider": "Google",
      "rank": 11,
      "score": 101.4,
      "stabilityRange": [
        91.2,
        111.5
      ],
      "provisional": false,
      "benchmarksUsed": 41,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": 0.1233,
          "benchmarksObserved": 9
        },
        {
          "domainId": "coding_swe",
          "score": -0.1001,
          "benchmarksObserved": 11
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.0392,
          "benchmarksObserved": 2
        },
        {
          "domainId": "professional_work",
          "score": -0.0862,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": 0.246,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": 0.0695,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": null,
          "benchmarksObserved": 0
        }
      ]
    },
    {
      "modelId": "gpt-5-6-terra",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gpt-5-6-terra",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gpt-5-6-terra",
      "name": "GPT-5.6 Terra",
      "provider": "OpenAI",
      "rank": 12,
      "score": 95.1,
      "stabilityRange": [
        88.8,
        101.3
      ],
      "provisional": false,
      "benchmarksUsed": 54,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.1199,
          "benchmarksObserved": 13
        },
        {
          "domainId": "coding_swe",
          "score": -0.1159,
          "benchmarksObserved": 13
        },
        {
          "domainId": "agentic_tooluse",
          "score": 0.1187,
          "benchmarksObserved": 6
        },
        {
          "domainId": "professional_work",
          "score": -0.2336,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": -0.2919,
          "benchmarksObserved": 5
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.2134,
          "benchmarksObserved": 3
        }
      ]
    },
    {
      "modelId": "qwen-3-8-max",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/qwen-3-8-max",
      "currentEvidenceUrl": "https://thelatent.co/score/models/qwen-3-8-max",
      "name": "Qwen3.8-Max",
      "provider": "Alibaba",
      "rank": 13,
      "score": 94.6,
      "stabilityRange": [
        84.4,
        105
      ],
      "provisional": false,
      "benchmarksUsed": 43,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.1954,
          "benchmarksObserved": 7
        },
        {
          "domainId": "coding_swe",
          "score": -0.2515,
          "benchmarksObserved": 9
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.171,
          "benchmarksObserved": 5
        },
        {
          "domainId": "professional_work",
          "score": -0.2384,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": 0.0839,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": -0.1753,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": -0.1132,
          "benchmarksObserved": 3
        }
      ]
    },
    {
      "modelId": "muse-spark-1-2",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/muse-spark-1-2",
      "currentEvidenceUrl": "https://thelatent.co/score/models/muse-spark-1-2",
      "name": "Muse Spark 1.2",
      "provider": "Meta",
      "rank": 14,
      "score": 92.2,
      "stabilityRange": [
        83.7,
        100.7
      ],
      "provisional": false,
      "benchmarksUsed": 36,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.1623,
          "benchmarksObserved": 7
        },
        {
          "domainId": "coding_swe",
          "score": -0.3142,
          "benchmarksObserved": 7
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.1895,
          "benchmarksObserved": 5
        },
        {
          "domainId": "professional_work",
          "score": -0.1607,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": -0.2095,
          "benchmarksObserved": 2
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.3042,
          "benchmarksObserved": 1
        }
      ]
    },
    {
      "modelId": "deepseek-v4-pro-max",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/deepseek-v4-pro-max",
      "currentEvidenceUrl": "https://thelatent.co/score/models/deepseek-v4-pro-max",
      "name": "DeepSeek V4 Pro",
      "provider": "DeepSeek",
      "rank": 15,
      "score": 83.2,
      "stabilityRange": [
        74.1,
        93.9
      ],
      "provisional": false,
      "benchmarksUsed": 35,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.5014,
          "benchmarksObserved": 10
        },
        {
          "domainId": "coding_swe",
          "score": -0.1788,
          "benchmarksObserved": 8
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.6159,
          "benchmarksObserved": 3
        },
        {
          "domainId": "professional_work",
          "score": -0.675,
          "benchmarksObserved": 13
        },
        {
          "domainId": "multimodal",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.4708,
          "benchmarksObserved": 1
        }
      ]
    },
    {
      "modelId": "gpt-5-6-luna",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/gpt-5-6-luna",
      "currentEvidenceUrl": "https://thelatent.co/score/models/gpt-5-6-luna",
      "name": "GPT-5.6 Luna",
      "provider": "OpenAI",
      "rank": 16,
      "score": 81.3,
      "stabilityRange": [
        73,
        90.1
      ],
      "provisional": false,
      "benchmarksUsed": 54,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.7319,
          "benchmarksObserved": 13
        },
        {
          "domainId": "coding_swe",
          "score": -0.4049,
          "benchmarksObserved": 12
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.4101,
          "benchmarksObserved": 7
        },
        {
          "domainId": "professional_work",
          "score": -0.4576,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": -0.536,
          "benchmarksObserved": 5
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.6222,
          "benchmarksObserved": 3
        }
      ]
    },
    {
      "modelId": "sonnet-5",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/sonnet-5",
      "currentEvidenceUrl": "https://thelatent.co/score/models/sonnet-5",
      "name": "Sonnet 5",
      "provider": "Anthropic",
      "rank": 17,
      "score": 81.3,
      "stabilityRange": [
        74.3,
        88.5
      ],
      "provisional": false,
      "benchmarksUsed": 51,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.4027,
          "benchmarksObserved": 10
        },
        {
          "domainId": "coding_swe",
          "score": -0.5434,
          "benchmarksObserved": 14
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.6853,
          "benchmarksObserved": 6
        },
        {
          "domainId": "professional_work",
          "score": -0.3787,
          "benchmarksObserved": 14
        },
        {
          "domainId": "multimodal",
          "score": -0.6721,
          "benchmarksObserved": 4
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.4748,
          "benchmarksObserved": 3
        }
      ]
    },
    {
      "modelId": "qwen-3-8-27b",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/qwen-3-8-27b",
      "currentEvidenceUrl": "https://thelatent.co/score/models/qwen-3-8-27b",
      "name": "Qwen3.8-27B",
      "provider": "Alibaba",
      "rank": 18,
      "score": 71,
      "stabilityRange": [
        61.8,
        79.9
      ],
      "provisional": false,
      "benchmarksUsed": 39,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -1.0717,
          "benchmarksObserved": 8
        },
        {
          "domainId": "coding_swe",
          "score": -1.0363,
          "benchmarksObserved": 8
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.6387,
          "benchmarksObserved": 3
        },
        {
          "domainId": "professional_work",
          "score": -0.7682,
          "benchmarksObserved": 12
        },
        {
          "domainId": "multimodal",
          "score": -0.4692,
          "benchmarksObserved": 6
        },
        {
          "domainId": "security",
          "score": null,
          "benchmarksObserved": 0
        },
        {
          "domainId": "preference_communication",
          "score": -0.9063,
          "benchmarksObserved": 2
        }
      ]
    },
    {
      "modelId": "deepseek-v4-flash",
      "evidenceUrl": "https://thelatent.co/score/releases/tli-2026-v1.0-2026-09-10-benchmarks-1/deepseek-v4-flash",
      "currentEvidenceUrl": "https://thelatent.co/score/models/deepseek-v4-flash",
      "name": "DeepSeek V4 Flash",
      "provider": "DeepSeek",
      "rank": 19,
      "score": 67.2,
      "stabilityRange": [
        59.6,
        74
      ],
      "provisional": false,
      "benchmarksUsed": 37,
      "domains": [
        {
          "domainId": "knowledge_reasoning",
          "score": -0.8907,
          "benchmarksObserved": 10
        },
        {
          "domainId": "coding_swe",
          "score": -0.8351,
          "benchmarksObserved": 9
        },
        {
          "domainId": "agentic_tooluse",
          "score": -0.7732,
          "benchmarksObserved": 5
        },
        {
          "domainId": "professional_work",
          "score": -0.9647,
          "benchmarksObserved": 10
        },
        {
          "domainId": "multimodal",
          "score": -0.9718,
          "benchmarksObserved": 1
        },
        {
          "domainId": "security",
          "score": -0.8704,
          "benchmarksObserved": 1
        },
        {
          "domainId": "preference_communication",
          "score": -1.0048,
          "benchmarksObserved": 1
        }
      ]
    }
  ]
}