{
  "api_version": "1",
  "updated": "2026-11-10",
  "latest_event": "2026-11-09",
  "rights": "© Frontier Log, all rights reserved.",
  "benchmark": {
    "slug": "hle",
    "name": "Humanity's Last Exam",
    "family": "hle",
    "unit": "%",
    "higher_is_better": true,
    "status": "active",
    "active": true,
    "superseded_by": null,
    "retired": null,
    "benchmark_url": "https://lastexam.ai",
    "introduced": "2025-01-23",
    "creator": "Center for AI Safety and Scale AI",
    "introduced_source": "https://scale.com/blog/humanitys-last-exam-results",
    "settings": [
      {
        "setting": "no-tools",
        "name": "No tools, full set",
        "url": "https://frontierlog.ai/benchmark/hle",
        "results": 38
      },
      {
        "setting": "with-tools",
        "name": "With tools, full set",
        "url": "https://frontierlog.ai/benchmark/hle/with-tools",
        "results": 41
      },
      {
        "setting": "no-tools-text-only",
        "name": "No tools, text-only subset",
        "url": "https://frontierlog.ai/benchmark/hle/no-tools-text-only",
        "results": 29
      },
      {
        "setting": "with-tools-text-only",
        "name": "With tools, text-only subset",
        "url": "https://frontierlog.ai/benchmark/hle/with-tools-text-only",
        "results": 15
      },
      {
        "setting": "no-tools-unstated",
        "name": "No tools, set not stated",
        "url": "https://frontierlog.ai/benchmark/hle/no-tools-unstated",
        "results": 41
      },
      {
        "setting": "with-tools-unstated",
        "name": "With tools, set not stated",
        "url": "https://frontierlog.ai/benchmark/hle/with-tools-unstated",
        "results": 28
      }
    ]
  },
  "total": 28,
  "results": [
    {
      "id": "openai-o3:hle:with-tools-unstated:1",
      "model": "openai-o3",
      "model_name": "OpenAI o3",
      "model_date": "2025-04-16",
      "event": "https://frontierlog.ai/event/2025-04-16-o3-and-o4-mini",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 24.9,
      "unit": "%",
      "printed": "24.90",
      "reported_by": "openai",
      "run_by": null,
      "label": "o3 (python + browsing** tools)",
      "source": "https://openai.com/index/introducing-o3-and-o4-mini",
      "source_title": "Introducing OpenAI o3 and o4-mini",
      "source_date": "2025-04-16",
      "where": "'What's changed' section, bar chart 'Humanity's Last Exam', 4th bar 'o3 (python + browsing** tools)' (page screenshot (job 20879 capture), part 02)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "high reasoning effort; Python and browsing; known cheating domains blocked; attempts a monitor flags count as wrong",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://openai.com/index/introducing-gpt-oss",
          "name": "gpt-oss-120b and gpt-oss-20b",
          "title": "Introducing gpt-oss",
          "lab": "openai",
          "date": "2025-08-05",
          "printed": "24.9",
          "where": "Humanity's Last Exam chart, bar 'o3 (with tools)'"
        },
        {
          "url": "https://openai.com/index/introducing-chatgpt-agent",
          "name": "OpenAI launches ChatGPT agent",
          "title": "Introducing ChatGPT agent",
          "lab": "openai",
          "date": "2025-07-17",
          "printed": "24.9%",
          "where": "Humanity's Last Exam (Full Set) chart, accuracy pass@1, bar 'OpenAI o3 (python + browsing)'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/392f30ae76e5713f.png"
    },
    {
      "id": "openai-o4-mini:hle:with-tools-unstated:1",
      "model": "openai-o4-mini",
      "model_name": "o4-mini",
      "model_date": "2025-04-16",
      "event": "https://frontierlog.ai/event/2025-04-16-o3-and-o4-mini",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 17.7,
      "unit": "%",
      "printed": "17.70",
      "reported_by": "openai",
      "run_by": null,
      "label": "o4-mini (with python + browsing** tools)",
      "source": "https://openai.com/index/introducing-o3-and-o4-mini",
      "source_title": "Introducing OpenAI o3 and o4-mini",
      "source_date": "2025-04-16",
      "where": "'What's changed' section, bar chart 'Humanity's Last Exam', 6th bar 'o4-mini (with python + browsing** tools)' (page screenshot (job 20879 capture), part 02)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "high reasoning effort; Python and browsing; known cheating domains blocked; attempts a monitor flags count as wrong",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-26",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/392f30ae76e5713f.png"
    },
    {
      "id": "gpt-oss-120b:hle:with-tools-unstated:1",
      "model": "gpt-oss-120b",
      "model_name": "gpt-oss-120b",
      "model_date": "2025-08-05",
      "event": "https://frontierlog.ai/event/2025-08-05-gpt-oss",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 19.0,
      "unit": "%",
      "printed": "19",
      "reported_by": "openai",
      "run_by": null,
      "label": "gpt-oss-120b (with tools)",
      "source": "https://openai.com/index/introducing-gpt-oss",
      "source_title": "Introducing gpt-oss",
      "source_date": "2025-08-05",
      "where": "'Evaluations' section, bar chart 'Humanity's Last Exam / Expert-level questions across subjects', 1st bar 'gpt-oss-120b (with tools)' (page screenshot (job 20879 capture), part 02)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "tools not named",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/ec5cfc228864ff6c.png"
    },
    {
      "id": "gpt-5-mini:hle:with-tools-unstated:1",
      "model": "gpt-5-mini",
      "model_name": "GPT-5 mini",
      "model_date": "2025-08-07",
      "event": "https://frontierlog.ai/event/2025-08-07-gpt-5",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 31.6,
      "unit": "%",
      "printed": "31.6%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5 mini (high)",
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano",
      "source_title": "Introducing GPT-5.4 mini and nano",
      "source_date": "2026-03-17",
      "where": "detailed results, 'Intelligence' table (HTML; columns GPT-5.4 (xhigh), GPT-5.4 mini (xhigh), GPT-5.4 nano (xhigh), GPT-5 mini (high¹)), row 'HLE w/ tool', column 'GPT-5 mini (high¹)'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "high reasoning effort, the highest available; tools not named",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-11-07",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/6dd7cf7ff7e77d25.png"
    },
    {
      "id": "glm-4-6:hle:with-tools-unstated:1",
      "model": "glm-4-6",
      "model_name": "GLM-4.6",
      "model_date": "2025-09-30",
      "event": "https://frontierlog.ai/event/2025-09-30-glm-4-6",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 30.4,
      "unit": "%",
      "printed": "30.4",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-4.6",
      "source": "https://z.ai/blog/glm-4.6",
      "source_title": "GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilities",
      "source_date": "2025-09-30",
      "where": "benchmark chart image (eight panels), panel 'HLE', first bar 'GLM-4.6', label above the bar marked 'w/ Tools'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "128K context",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://z.ai/blog/glm-4.7",
          "name": "GLM-4.7",
          "title": "GLM-4.7: Advancing the Coding Capability",
          "lab": "zai",
          "date": "2025-12-22",
          "printed": "30.4",
          "where": "Benchmark table, row 'HLE w/ Tools', column 'GLM-4.6'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/e2ec892bfba7273b.png",
      "source_image_original": "https://z-cdn.chatglm.cn/z-blog/glm-4-6/coding_benchmark.png"
    },
    {
      "id": "gpt-5-1:hle:with-tools-unstated:1",
      "model": "gpt-5-1",
      "model_name": "GPT-5.1",
      "model_date": "2025-11-12",
      "event": "https://frontierlog.ai/event/2025-11-12-gpt-5-1",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 42.7,
      "unit": "%",
      "printed": "42.7%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5.1 Thinking",
      "source": "https://openai.com/index/introducing-gpt-5-2",
      "source_title": "Introducing GPT-5.2",
      "source_date": "2025-12-11",
      "where": "science and math eval table, column 'GPT-5.1 Thinking', row 'HLE (w/ search, Python)'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "search and Python",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://z.ai/blog/glm-4.7",
          "name": "GLM-4.7",
          "title": "GLM-4.7: Advancing the Coding Capability",
          "lab": "zai",
          "date": "2025-12-22",
          "printed": "42.7",
          "where": "Benchmark table, row 'HLE w/ Tools', column 'GPT-5.1 High'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
    },
    {
      "id": "gpt-5-2:hle:with-tools-unstated:1",
      "model": "gpt-5-2",
      "model_name": "GPT-5.2",
      "model_date": "2025-12-11",
      "event": "https://frontierlog.ai/event/2025-12-11-gpt-5-2",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 45.5,
      "unit": "%",
      "printed": "45.5%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5.2 Thinking",
      "source": "https://openai.com/index/introducing-gpt-5-2",
      "source_title": "Introducing GPT-5.2",
      "source_date": "2025-12-11",
      "where": "science and math eval table, column 'GPT-5.2 Thinking', row 'HLE (w/ search, Python)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "search and Python",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-4-6-system-card",
          "name": "Claude Opus 4.6 card",
          "title": "Claude Opus 4.6 System Card",
          "lab": "anthropic",
          "date": "2026-02-05",
          "printed": "45.5%",
          "where": "Figure 2.21.2.A (p.42), right panel With Tools, bar GPT-5.2 Thinking value label"
        },
        {
          "url": "https://www.anthropic.com/news/claude-opus-4-6",
          "name": "Claude Opus 4.6",
          "title": "Claude Opus 4.6",
          "lab": "anthropic",
          "date": "2026-02-05",
          "printed": "45.5",
          "where": "Bar chart 'Multidisciplinary reasoning, Humanity's Last Exam', bar 'GPT-5.2', darker 'With tools' bar"
        },
        {
          "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/",
          "name": "Gemini 3.1 Pro",
          "title": "Gemini 3.1 Pro: A smarter model for your most complex tasks",
          "lab": "google",
          "date": "2026-02-19",
          "printed": "45.5%",
          "where": "Benchmark table image; row 'Humanity's Last Exam, Search (blocklist) + Code'; column gpt-5-2"
        },
        {
          "url": "https://www.kimi.ai/blog/kimi-k2-5",
          "name": "Moonshot AI launches Kimi Code",
          "title": "Kimi K2.5: Visual Agentic Intelligence",
          "lab": "moonshot",
          "date": "2026-01-27",
          "printed": "45.5",
          "where": "Appendix Benchmark table, row HLE-Full w/ tools ; column GPT-5.2 (xhigh)"
        },
        {
          "url": "https://openai.com/index/introducing-gpt-5-4",
          "name": "GPT-5.4",
          "title": "Introducing GPT-5.4",
          "lab": "openai",
          "date": "2026-03-05",
          "printed": "45.5%",
          "where": "table 6 \"Humanity's Last Exam (with tools)\" row, \"GPT-5.2\" column"
        },
        {
          "url": "https://qwen.ai/blog?id=qwen3-max-thinking",
          "name": "Qwen3-Max-Thinking",
          "title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
          "lab": "alibaba",
          "date": "2026-01-26",
          "printed": "45.5",
          "where": "Score chart image (score.png), panel Humanity's Last Exam (with Search), value label on the GPT-5.2 bar"
        },
        {
          "url": "https://z.ai/blog/glm-5",
          "name": "GLM-5",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "45.5",
          "where": "Benchmark table, row 'Humanity's Last Exam w/ Tools (* = full set for Opus 4.5, Gemini, GPT-5.2)', column 'GPT-5.2 (xhigh)'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/22817d0b2b0a2432.png"
    },
    {
      "id": "nemotron-3-nano-30b-a3b:hle:with-tools-unstated:1",
      "model": "nemotron-3-nano-30b-a3b",
      "model_name": "Nemotron 3 Nano 30B-A3B",
      "model_date": "2025-12-15",
      "event": "https://frontierlog.ai/event/2025-12-15-nemotron-3",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 15.5,
      "unit": "%",
      "printed": "15.5",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8/blob/fa5699ff0bb9da723217af52488c0f9653512ffb/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model card (launch-day revision)",
      "source_date": "2025-12-15",
      "where": "'Reasoning Benchmark Evaluations' table, row 'HLE (with tools)', column 'NVIDIA-Nemotron-3-Nano-30B-A3B-BF16'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Nano-Technical-Report.pdf",
          "name": "Nemotron 3 card",
          "title": "Nemotron 3 Nano: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning",
          "lab": "nvidia",
          "date": "2025-12-15",
          "printed": "15.48",
          "where": "Table 3, HLE (with tools) row, N-3-Nano column"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/blob/bf77c3174f68ad409e1c2aa60daeb46e32d1c606/README.md",
          "name": "Nemotron 3 card",
          "title": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2025-12-15",
          "printed": "15.5",
          "where": "Benchmarks table, 'Nemotron-3-Nano' (first, BF16) column, HLE (with tools) row"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/4ff9ee8b68e61bc2.png"
    },
    {
      "id": "glm-4-7:hle:with-tools-unstated:1",
      "model": "glm-4-7",
      "model_name": "GLM-4.7",
      "model_date": "2025-12-22",
      "event": "https://frontierlog.ai/event/2025-12-22-glm-4-7",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 42.8,
      "unit": "%",
      "printed": "42.8",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-4.7",
      "source": "https://z.ai/blog/glm-4.7",
      "source_title": "GLM-4.7: Advancing the Coding Capability",
      "source_date": "2025-12-22",
      "where": "benchmark table, row 'HLE w/ Tools', column 'GLM-4.7'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://z.ai/blog/glm-5",
          "name": "GLM-5",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "42.8",
          "where": "Benchmark table, row 'Humanity's Last Exam w/ Tools (* = full set for Opus 4.5, Gemini, GPT-5.2)', column 'GLM-4.7 (Thinking)'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/b52f90e840f8cbd6.png"
    },
    {
      "id": "qwen3-max-thinking:hle:with-tools-unstated:1",
      "model": "qwen3-max-thinking",
      "model_name": "Qwen3-Max-Thinking",
      "model_date": "2026-01-26",
      "event": "https://frontierlog.ai/event/2026-01-26-qwen3-max-thinking",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 49.8,
      "unit": "%",
      "printed": "49.8",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3-Max-Thinking",
      "source": "https://qwen.ai/blog?id=qwen3-max-thinking",
      "source_title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
      "source_date": "2026-01-26",
      "where": "evaluation table, row 'HLE (w/ tools)', column 'Qwen3-Max-Thinking'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "with search, HLE-related websites blocked",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/aed8db26403e0575.png"
    },
    {
      "id": "qwen3-max-thinking:hle:with-tools-unstated:2",
      "model": "qwen3-max-thinking",
      "model_name": "Qwen3-Max-Thinking",
      "model_date": "2026-01-26",
      "event": "https://frontierlog.ai/event/2026-01-26-qwen3-max-thinking",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 58.3,
      "unit": "%",
      "printed": "58.3",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3-Max-Thinking (with TTS)",
      "source": "https://qwen.ai/blog?id=qwen3-max-thinking",
      "source_title": "Pushing Qwen3-Max-Thinking Beyond its Limits",
      "source_date": "2026-01-26",
      "where": "score chart image (seven panels, below the banner), panel 'Humanity's Last Exam (with Search)', first bar, upper value label (legend 'Qwen3-Max-Thinking (with TTS)')",
      "own": true,
      "primary": false,
      "parallel": true,
      "notes": "heavy mode: test-time scaling over parallel trajectories; with search",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/80d8b6e334ec18d8.png",
      "source_image_original": "https://qianwen-res.oss-accelerate-overseas.aliyuncs.com/Qwen3-Max-Thinking/score.png"
    },
    {
      "id": "qwen3-5-397b-a17b:hle:with-tools-unstated:1",
      "model": "qwen3-5-397b-a17b",
      "model_name": "Qwen3.5-397B-A17B",
      "model_date": "2026-02-16",
      "event": "https://frontierlog.ai/event/2026-02-16-qwen3-5",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 48.3,
      "unit": "%",
      "printed": "48.3",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.5-397B-A17B",
      "source": "https://qwen.ai/blog?id=qwen3.5",
      "source_title": "Qwen3.5: Towards Native Multimodal Agents",
      "source_date": "2026-02-16",
      "where": "'Performance' section, 'Language' table, row 'HLE w/ tool', column 'Qwen3.5-397B-A17B'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "search agent; context folding at 256K tokens",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/a2e98da83152d308.png"
    },
    {
      "id": "gpt-5-4:hle:with-tools-unstated:1",
      "model": "gpt-5-4",
      "model_name": "GPT-5.4",
      "model_date": "2026-03-05",
      "event": "https://frontierlog.ai/event/2026-03-05-gpt-5-4",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 52.1,
      "unit": "%",
      "printed": "52.1%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.4",
      "source": "https://openai.com/index/introducing-gpt-5-4",
      "source_title": "Introducing GPT-5.4",
      "source_date": "2026-03-05",
      "where": "academic eval table, column 'GPT‑5.4', row 'Humanity's Last Exam (with tools)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-4-7-system-card",
          "name": "Claude Opus 4.7 card",
          "title": "Claude Opus 4.7 System Card",
          "lab": "anthropic",
          "date": "2026-04-16",
          "printed": "52.1%",
          "where": "Table 8.1.A (p.191), row Humanity's Last Exam / With tools, column GPT-5.4"
        },
        {
          "url": "https://www.kimi.ai/blog/kimi-k2-6",
          "name": "Kimi K2.6",
          "title": "Kimi K2.6: Advancing Open-Source Coding",
          "lab": "moonshot",
          "date": "2026-04-20",
          "printed": "52.1",
          "where": "Benchmark Table (appendix), row HLE-Full w/ tools ; column GPT-5.4 (xhigh)"
        },
        {
          "url": "https://openai.com/index/introducing-gpt-5-5",
          "name": "GPT-5.5",
          "title": "Introducing GPT-5.5",
          "lab": "openai",
          "date": "2026-04-23",
          "printed": "52.1%",
          "where": "table 5 \"Humanity's Last Exam (with tools)\" row, \"GPT-5.4\" column"
        },
        {
          "url": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano",
          "name": "GPT-5.4 mini and nano",
          "title": "Introducing GPT-5.4 mini and nano",
          "lab": "openai",
          "date": "2026-03-17",
          "printed": "52.1%",
          "where": "table 3 \"HLE w/ tool\" row, \"GPT-5.4 (xhigh)\" column"
        },
        {
          "url": "https://z.ai/blog/glm-5.1",
          "name": "GLM-5.1",
          "title": "GLM-5.1: Towards Long-Horizon Tasks",
          "lab": "zai",
          "date": "2026-04-07",
          "printed": "52.1",
          "where": "Benchmark table, row 'HLE w/ Tools (* = full set for Opus 4.6, Gemini 3.1 Pro, GPT-5.4)', column 'GPT-5.4'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/2322d7fe4c245f75.png"
    },
    {
      "id": "nemotron-3-super-120b-a12b:hle:with-tools-unstated:1",
      "model": "nemotron-3-super-120b-a12b",
      "model_name": "Nemotron 3 Super 120B-A12B",
      "model_date": "2026-03-11",
      "event": "https://frontierlog.ai/event/2026-03-11-nemotron-3-super",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 22.8,
      "unit": "%",
      "printed": "22.8",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron-3-Super-120B-A12B-BF16",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8/blob/1408abd63ba6ba1c37a78b94d332dac35da45370/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8 model card (launch-day revision)",
      "source_date": "2026-03-11",
      "where": "accuracy chart (image) at the top of the card, group 'HLE (Science)', 'Nemotron-3-Super-120B-A12B-BF16' bar (dark green, first), upper label marked '+tools'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Super-Technical-Report.pdf",
          "name": "Nemotron 3 Super card",
          "title": "Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "22.82",
          "where": "Table 5, HLE (with tools) row, N-3-Super column"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16/blob/2dc98e2afe4face0e4ce40972a915c45368bd34a/README.md",
          "name": "Nemotron 3 Super card",
          "title": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16 model card (revision of 2026-08-25)",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "22.82",
          "where": "Benchmarks table, 'Nemotron 3 Super' column, HLE (with tools) row"
        },
        {
          "url": "https://build.nvidia.com/nvidia/nemotron-3-super-120b-a12b/modelcard",
          "name": "Nemotron 3 Super card",
          "title": "nemotron-3-super-120b-a12b model card (NVIDIA NIM)",
          "lab": "nvidia",
          "date": "2026-03-11",
          "printed": "22.82",
          "where": "Benchmarks table, HLE (with tools) row, Nemotron 3 Super column"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/6fd72a7dcf869cac.png",
      "source_image_original": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8/resolve/1408abd63ba6ba1c37a78b94d332dac35da45370/accuracy_chart.png"
    },
    {
      "id": "gpt-5-4-mini:hle:with-tools-unstated:1",
      "model": "gpt-5-4-mini",
      "model_name": "GPT-5.4 mini",
      "model_date": "2026-03-17",
      "event": "https://frontierlog.ai/event/2026-03-17-gpt-5-4-mini-nano",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 41.5,
      "unit": "%",
      "printed": "41.5%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5.4 mini (xhigh)",
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano",
      "source_title": "Introducing GPT-5.4 mini and nano",
      "source_date": "2026-03-17",
      "where": "academic table (columns GPT-5.4 (xhigh), GPT-5.4 mini (xhigh), ...), row 'HLE w/ tool', column 'GPT-5.4 mini (xhigh)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "xhigh effort; tools not named",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/ec1c7aef5bba5db5.png"
    },
    {
      "id": "gpt-5-4-nano:hle:with-tools-unstated:1",
      "model": "gpt-5-4-nano",
      "model_name": "GPT-5.4 nano",
      "model_date": "2026-03-17",
      "event": "https://frontierlog.ai/event/2026-03-17-gpt-5-4-mini-nano",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 37.7,
      "unit": "%",
      "printed": "37.7%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5.4 nano (xhigh)",
      "source": "https://openai.com/index/introducing-gpt-5-4-mini-and-nano",
      "source_title": "Introducing GPT-5.4 mini and nano",
      "source_date": "2026-03-17",
      "where": "academic table (columns GPT-5.4 (xhigh), GPT-5.4 mini (xhigh), GPT-5.4 nano (xhigh), ...), row 'HLE w/ tool', column 'GPT-5.4 nano (xhigh)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "xhigh effort; tools not named",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-26",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/ec1c7aef5bba5db5.png"
    },
    {
      "id": "qwen3-6-plus:hle:with-tools-unstated:1",
      "model": "qwen3-6-plus",
      "model_name": "Qwen3.6-Plus",
      "model_date": "2026-04-02",
      "event": "https://frontierlog.ai/event/2026-04-02-qwen3-6-plus",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 50.6,
      "unit": "%",
      "printed": "50.6",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.6-Plus",
      "source": "https://qwen.ai/blog?id=qwen3.6",
      "source_title": "Qwen3.6-Plus: Towards Real World Agents",
      "source_date": "2026-04-02",
      "where": "'Performance' section, 'Language' table, row 'HLE w/ tool', column 'Qwen3.6-Plus'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "256K context with context folding",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://z.ai/blog/glm-5.1",
          "name": "GLM-5.1",
          "title": "GLM-5.1: Towards Long-Horizon Tasks",
          "lab": "zai",
          "date": "2026-04-07",
          "printed": "50.6",
          "where": "Benchmark table, row 'HLE w/ Tools (* = full set for Opus 4.6, Gemini 3.1 Pro, GPT-5.4)', column 'Qwen3.6-Plus'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/6f4855b376c670b4.png"
    },
    {
      "id": "qwen3-6-plus:hle:with-tools-unstated:2",
      "model": "qwen3-6-plus",
      "model_name": "Qwen3.6-Plus",
      "model_date": "2026-04-02",
      "event": "https://frontierlog.ai/event/2026-04-02-qwen3-6-plus",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 50.2,
      "unit": "%",
      "printed": "50.2",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.6-Plus",
      "source": "https://qwen.ai/blog?id=qwen3.7",
      "source_title": "Qwen3.7: The Agent Frontier",
      "source_date": "2026-05-20",
      "where": "Qwen3.7 post, 'Performance' table, row 'HLE w/ tools', column 'Qwen3.6-Plus'",
      "own": false,
      "primary": false,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_capture": "https://frontierlog.ai/captures/9d2174c2d8ef254a.png"
    },
    {
      "id": "gpt-5-5:hle:with-tools-unstated:1",
      "model": "gpt-5-5",
      "model_name": "GPT-5.5",
      "model_date": "2026-04-23",
      "event": "https://frontierlog.ai/event/2026-04-23-gpt-5-5",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 52.2,
      "unit": "%",
      "printed": "52.2%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.5",
      "source": "https://openai.com/index/introducing-gpt-5-5",
      "source_title": "Introducing GPT-5.5",
      "source_date": "2026-04-23",
      "where": "academic eval table, column 'GPT‑5.5', row 'Humanity's Last Exam (with tools)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
          "name": "Claude Fable 5 and Mythos 5 card",
          "title": "Claude Fable 5 and Claude Mythos 5 System Card",
          "lab": "anthropic",
          "date": "2026-06-09",
          "printed": "52.2",
          "where": "Table 8.1.A (p.251), row Humanity's Last Exam / With tools, column GPT-5.5"
        },
        {
          "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
          "name": "Claude Opus 4.8 card",
          "title": "Claude Opus 4.8 System Card",
          "lab": "anthropic",
          "date": "2026-05-28",
          "printed": "52.2",
          "where": "Table 8.1.A (p.194), row Humanity's Last Exam / With tools, column GPT-5.5"
        },
        {
          "url": "https://www.anthropic.com/claude-sonnet-5-system-card",
          "name": "Claude Sonnet 5 card",
          "title": "Claude Sonnet 5 System Card",
          "lab": "anthropic",
          "date": "2026-06-30",
          "printed": "52.2",
          "where": "Table 8.1.A (p.115), row Humanity's Last Exam / With tools, column GPT-5.5"
        },
        {
          "url": "https://www.anthropic.com/news/claude-opus-4-8",
          "name": "Claude Opus 4.8",
          "title": "Introducing Claude Opus 4.8",
          "lab": "anthropic",
          "date": "2026-05-28",
          "printed": "52.2%",
          "where": "benchmark table image, row Humanity's Last Exam (with tools), column GPT-5.5"
        },
        {
          "url": "https://www.kimi.ai/blog/kimi-k3",
          "name": "Kimi K3",
          "title": "Kimi K3: Open Frontier Intelligence",
          "lab": "moonshot",
          "date": "2026-07-16",
          "printed": "52.2",
          "where": "Full Benchmark Table, row HLE-Full w/ tools ; column GPT 5.5 (xhigh)"
        },
        {
          "url": "https://z.ai/blog/glm-5.2",
          "name": "GLM-5.2",
          "title": "GLM-5.2: Built for Long-Horizon Tasks",
          "lab": "zai",
          "date": "2026-06-16",
          "printed": "52.2",
          "where": "Full Benchmark Table, row 'HLE w/ Tools (* = full set for Opus 4.8, GPT-5.5, Gemini 3.1 Pro)', column 'GPT-5.5'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/4fc1be3b62dac325.png"
    },
    {
      "id": "deepseek-v4-pro:hle:with-tools-unstated:1",
      "model": "deepseek-v4-pro",
      "model_name": "DeepSeek-V4-Pro",
      "model_date": "2026-04-24",
      "event": "https://frontierlog.ai/event/2026-04-24-deepseek-v4-preview",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 48.2,
      "unit": "%",
      "printed": "48.2",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DS-V4-Pro Max",
      "source": "https://api-docs.deepseek.com/news/news260424/",
      "source_title": "DeepSeek V4 Preview Release",
      "source_date": "2026-04-24",
      "where": "benchmark table image (v4-benchmark-2), section 'Agentic', row 'HLE w/tools (Pass@1)', column 'DS-V4-Pro Max'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "max reasoning effort",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://api-docs.deepseek.com/news/news260813/",
          "name": "DeepSeek-V4-Pro GA",
          "title": "DeepSeek-V4-Pro GA Release",
          "lab": "deepseek",
          "date": "2026-08-13",
          "printed": "48.2",
          "where": "Benchmark table image, row 'HLE (wo/w tools)', column 'DeepSeek-V4-Pro-Preview', printed '37.7/48.2'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/9226203de8e24b33.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4-benchmark-2.png"
    },
    {
      "id": "qwen3-7-max:hle:with-tools-unstated:1",
      "model": "qwen3-7-max",
      "model_name": "Qwen3.7-Max",
      "model_date": "2026-05-20",
      "event": "https://frontierlog.ai/event/2026-05-20-qwen3-7-max",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 53.5,
      "unit": "%",
      "printed": "53.5",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.7-Max",
      "source": "https://qwen.ai/blog?id=qwen3.7",
      "source_title": "Qwen3.7: The Agent Frontier",
      "source_date": "2026-05-20",
      "where": "'Performance' table, row 'HLE w/ tools', column 'Qwen3.7-Max'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://z.ai/blog/glm-5.2",
          "name": "GLM-5.2",
          "title": "GLM-5.2: Built for Long-Horizon Tasks",
          "lab": "zai",
          "date": "2026-06-16",
          "printed": "53.5",
          "where": "Full Benchmark Table, row 'HLE w/ Tools (* = full set for Opus 4.8, GPT-5.5, Gemini 3.1 Pro)', column 'Qwen3.7-Max'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/9d2174c2d8ef254a.png"
    },
    {
      "id": "nemotron-3-ultra-550b-a55b:hle:with-tools-unstated:1",
      "model": "nemotron-3-ultra-550b-a55b",
      "model_name": "Nemotron 3 Ultra 550B-A55B",
      "model_date": "2026-06-04",
      "event": "https://frontierlog.ai/event/2026-06-04-nemotron-3-ultra",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 37.4,
      "unit": "%",
      "printed": "37.4",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "N-3-Ultra",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Ultra-Technical-Report.pdf",
      "source_title": "Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning",
      "source_date": "2026-06-04",
      "where": "Table 10 (PDF page 38), row 'HLE (with tools)', column 'N-3-Ultra'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "37.40",
          "where": "Table 10 (Evaluation suite for Nemotron 3.5 Super), HLE (with tools, text) row, N3-Ultra 550B-A55B column"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16/blob/77df655d5e9f8362164ed14dd8b48f8bce657498/README.md",
          "name": "Nemotron 3 Ultra card",
          "title": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2026-06-04",
          "printed": "37.4",
          "where": "Benchmarks table, 'N-3-Ultra 550B-A55B' column, HLE (with tools) row"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/a737368a281c4576.png"
    },
    {
      "id": "inkling:hle:with-tools-unstated:1",
      "model": "inkling",
      "model_name": "Inkling",
      "model_date": "2026-07-15",
      "event": "https://frontierlog.ai/event/2026-07-15-inkling",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 46.0,
      "unit": "%",
      "printed": "46.0%",
      "reported_by": "thinking-machines",
      "run_by": null,
      "label": "Inkling effort=0.99",
      "source": "https://thinkingmachines.ai/news/introducing-inkling/",
      "source_title": "Inkling: Our Open-Weights Model",
      "source_date": "2026-07-15",
      "where": "section 'Benchmarking Inkling', main benchmark table, row 'HLE' with subtitle 'with tools' (Reasoning), column 'Inkling effort=0.99'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "effort 0.99; tools not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://thinkingmachines.ai/news/inkling-small/",
          "name": "Thinking Machines post",
          "title": "Introducing Inkling-Small",
          "lab": "thinking-machines",
          "date": "2026-07-30",
          "printed": "46.0%",
          "where": "Full eval suite table, row 'HLE† with tools', column 'Inkling'"
        },
        {
          "url": "https://huggingface.co/thinkingmachines/Inkling",
          "name": "Inkling card",
          "title": "thinkingmachines/Inkling model card",
          "lab": "thinking-machines",
          "date": "2026-07-15",
          "printed": "46.0%",
          "where": "Evaluations table (effort=0.99), row 'HLE (with tools)', column 'Inkling'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/846bbe93fde53e98.png"
    },
    {
      "id": "qwen3-8-max:hle:with-tools-unstated:1",
      "model": "qwen3-8-max",
      "model_name": "Qwen3.8-Max",
      "model_date": "2026-08-03",
      "event": "https://frontierlog.ai/event/2026-08-03-qwen3-8-max",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 56.2,
      "unit": "%",
      "printed": "56.2",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.8-Max",
      "source": "https://qwen.ai/blog?id=qwen3.8",
      "source_title": "Qwen3.8-Max: A New Bar for Coding and Cowork",
      "source_date": "2026-08-03",
      "where": "'Full Benchmark Table' (first, language/agent HTML table), General Agent section, row 'HLE w/ tools', column 'Qwen3.8-Max'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "tools not stated",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/zai-org/GLM-5.3",
          "name": "GLM-5.3 card",
          "title": "zai-org/GLM-5.3 model card",
          "lab": "zai",
          "date": "2026-08-14",
          "printed": "56.2",
          "where": "README Benchmark table, row 'HLE w/ Tools', column 'Qwen3.8-Max'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/10d02db2e5d51286.png"
    },
    {
      "id": "deepseek-v4-pro-0813:hle:with-tools-unstated:1",
      "model": "deepseek-v4-pro-0813",
      "model_name": "DeepSeek-V4-Pro-0813",
      "model_date": "2026-08-13",
      "event": "https://frontierlog.ai/event/2026-08-13-deepseek-v4-pro-ga",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 60.0,
      "unit": "%",
      "printed": "60.0",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V4-Pro-0813",
      "source": "https://api-docs.deepseek.com/news/news260813/",
      "source_title": "DeepSeek-V4-Pro GA Release",
      "source_date": "2026-08-13",
      "where": "table image, row 'HLE (wo/w tools)', column 'DeepSeek-V4-Pro-0813', second of the two values",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://api-docs.deepseek.com/news/news260910/",
          "name": "DeepSeek-V4.1-Flash",
          "title": "DeepSeek-V4.1-Flash: Smarter, Faster, More Efficient",
          "lab": "deepseek",
          "date": "2026-09-10",
          "printed": "60.0",
          "where": "Benchmark table image, row 'HLE (w/tools)', column 'DeepSeek V4-Pro 0813'"
        },
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash",
          "name": "DeepSeek-V4.1-Flash card",
          "title": "deepseek-ai/DeepSeek-V4.1-Flash model card",
          "lab": "deepseek",
          "date": "2026-09-10",
          "printed": "60.0",
          "where": "'Comparison with frontier models (Max reasoning effort)' table, row 'HLE w/ tools (Pass@1)', column 'DS-V4-Pro'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/51578c812ffc7347.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260813_benchmark_table_en.png"
    },
    {
      "id": "glm-5-3:hle:with-tools-unstated:1",
      "model": "glm-5-3",
      "model_name": "GLM-5.3",
      "model_date": "2026-08-14",
      "event": "https://frontierlog.ai/event/2026-08-14-glm-5-3",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 62.5,
      "unit": "%",
      "printed": "62.5",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.3",
      "source": "https://z.ai/blog/glm-5.3",
      "source_title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
      "source_date": "2026-08-14",
      "where": "'Performance across comparison models' table, row 'HLE w/ Tools', column 'GLM-5.3'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "300,000-token context with context management; GPT-5.6 Luna judge",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-19",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://api-docs.deepseek.com/news/news260910/",
          "name": "DeepSeek-V4.1-Flash",
          "title": "DeepSeek-V4.1-Flash: Smarter, Faster, More Efficient",
          "lab": "deepseek",
          "date": "2026-09-10",
          "printed": "62.5",
          "where": "Benchmark table image; row 'HLE (w/tools)'; column 'GLM 5.3'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/86a1ebfddd94c33e.png"
    },
    {
      "id": "gpt-6-astra:hle:with-tools-unstated:1",
      "model": "gpt-6-astra",
      "model_name": "GPT-6 Astra",
      "model_date": "2026-09-03",
      "event": "https://frontierlog.ai/event/2026-09-03-gpt-6-astra",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 57.2,
      "unit": "%",
      "printed": "57.2%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑6 Astra",
      "source": "https://openai.com/index/gpt-6-astra",
      "source_title": "GPT-6 Astra: A new generation of intelligence",
      "source_date": "2026-09-03",
      "where": "Academic table, row 'Humanity's Last Exam (w/ tools)', column 'GPT‑6 Astra'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-16",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-5-5-system-card",
          "name": "Claude Opus 5.5 card",
          "title": "Claude Opus 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-22",
          "printed": "57.2",
          "where": "Table 8.1.A (p.174), row Humanity's Last Exam / With tools, column GPT-6 Astra"
        },
        {
          "url": "https://www.anthropic.com/claude-opus-5-5",
          "name": "Claude Opus 5.5",
          "title": "Introducing Claude Opus 5.5",
          "lab": "anthropic",
          "date": "2026-09-22",
          "printed": "57.2%",
          "where": "HTML benchmark table under 'Performance and cost-effectiveness', row Humanity's Last Exam (with tools), column GPT-6 Astra"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/478a8e3e7a393b2d.png"
    },
    {
      "id": "deepseek-v4-1-flash:hle:with-tools-unstated:1",
      "model": "deepseek-v4-1-flash",
      "model_name": "DeepSeek-V4.1-Flash",
      "model_date": "2026-09-10",
      "event": "https://frontierlog.ai/event/2026-09-10-deepseek-v4-1-flash",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "with-tools-unstated",
      "setting_name": "With tools, set not stated",
      "score": 63.9,
      "unit": "%",
      "printed": "63.9",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek V4.1-Flash",
      "source": "https://api-docs.deepseek.com/news/news260910/",
      "source_title": "DeepSeek-V4.1-Flash: Smarter, Faster, More Efficient",
      "source_date": "2026-09-10",
      "where": "benchmark table image (second image), row 'HLE (w/tools)', column 'DeepSeek V4.1-Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash",
          "name": "DeepSeek-V4.1-Flash card",
          "title": "deepseek-ai/DeepSeek-V4.1-Flash model card",
          "lab": "deepseek",
          "date": "2026-09-10",
          "printed": "63.9",
          "where": "'Comparison with frontier models (Max reasoning effort)' table, row 'HLE w/ tools (Pass@1)', column 'DS-V4.1-Flash'"
        },
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/DeepSeek_V41_Tech_Report.pdf",
          "name": "DeepSeek-V4.1-Flash card",
          "title": "DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression",
          "lab": "deepseek",
          "date": "2026-09-10",
          "printed": "63.9",
          "where": "Instruct results table (Max reasoning effort), row 'HLE w/ tools (Pass@1)', column 'DS-V4.1-Flash'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/19d3b83b1b9e5f5b.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4.1_260910_benchmark_table_en.png"
    }
  ]
}