{
  "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": 29,
  "results": [
    {
      "id": "openai-o3-mini:hle:no-tools-text-only:1",
      "model": "openai-o3-mini",
      "model_name": "o3-mini",
      "model_date": "2025-01-31",
      "event": "https://frontierlog.ai/event/2025-01-31-openai-o3-mini",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 13.0,
      "unit": "%",
      "printed": "13.0",
      "reported_by": "openai",
      "run_by": null,
      "label": "OpenAI o3-mini (high)",
      "source": "https://openai.com/index/introducing-deep-research",
      "source_title": "Introducing deep research",
      "source_date": "2025-02-02",
      "where": "Humanity's Last Exam table (Model / Accuracy (%)), row 'OpenAI o3-mini (high)*'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "high reasoning effort",
      "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/9676fd3a951a3d1c.png"
    },
    {
      "id": "minimax-m1:hle:no-tools-text-only:1",
      "model": "minimax-m1",
      "model_name": "MiniMax-M1",
      "model_date": "2025-06-16",
      "event": "https://frontierlog.ai/event/2025-06-16-minimax-m1",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 8.4,
      "unit": "%",
      "printed": "8.4",
      "reported_by": "minimax",
      "run_by": null,
      "label": "MiniMax-M1-80k",
      "source": "https://www.minimax.io/news/minimaxm1",
      "source_title": "MiniMax-M1, the World's First Open-Source, Large-Scale, Hybrid-Attention Reasoning Model",
      "source_date": "2025-06-16",
      "where": "full results table image (last image, after the bullet list), row 'HLE (no tools)', column 'MiniMax-M1-80k'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "MiniMax-M1-80k, 80K thinking budget; text-only subset",
      "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/52957ae223f9c494.png",
      "source_image_original": "https://filecdn.minimax.chat/public/d50018df-d646-40e3-824b-e507c209ad03.png"
    },
    {
      "id": "qwen3-235b-a22b-thinking-2507:hle:no-tools-text-only:1",
      "model": "qwen3-235b-a22b-thinking-2507",
      "model_name": "Qwen3-235B-A22B-Thinking-2507",
      "model_date": "2025-07-25",
      "event": "https://frontierlog.ai/event/2025-07-25-qwen3-235b-a22b-thinking-2507",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 18.2,
      "unit": "%",
      "printed": "18.2",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3-235B-A22B-Thinking-2507",
      "source": "https://www.alizila.com/alibaba-unveils-new-qwen3-models-for-coding-complexing-reasoning-and-machine-translation/",
      "source_title": "Alibaba Unveils New Qwen3 Models for Coding, Complexing Reasoning and Machine Translation",
      "source_date": "2025-07-25",
      "where": "bar chart image in the 'Qwen3-235B' section, group 'HLE', first bar (legend 'Qwen3-235B-A22B-Thinking-2507') value label",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-26",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507/blob/4cd68849b2a0fadb84866c703b133aa8c8636130/README.md",
          "name": "Qwen3-235B-A22B-Thinking-2507 card",
          "title": "Qwen/Qwen3-235B-A22B-Thinking-2507 model card (revision of 2025-07-25)",
          "lab": "alibaba",
          "date": "2025-07-25",
          "printed": "18.2",
          "where": "Qwen3-235B-A22B-Thinking-2507 HF model card at revision 4cd68849, Performance table, row 'HLE', column 'Qwen3-235B-A22B-Thinking-2507'; printed '18.2#', footnote #: text-only subset"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/3ebde273acb680b6.png",
      "source_image_original": "https://alizila.oss-us-west-1.aliyuncs.com/uploads/2025/07/0CA37A90-CD03-4037-AF89-F1548BB1C156.png"
    },
    {
      "id": "glm-4-5:hle:no-tools-text-only:1",
      "model": "glm-4-5",
      "model_name": "GLM-4.5",
      "model_date": "2025-07-28",
      "event": "https://frontierlog.ai/event/2025-07-28-glm-4-5",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 14.4,
      "unit": "%",
      "printed": "14.4",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-4.5",
      "source": "https://z.ai/blog/glm-4.5",
      "source_title": "GLM-4.5: Reasoning, Coding, and Agentic Abililties",
      "source_date": "2025-07-28",
      "where": "'Reasoning' section table, row 'HLE', column 'GLM-4.5'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "thinking mode; text-only questions; GPT-4o judge",
      "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-4.6",
          "name": "GLM-4.6",
          "title": "GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilities",
          "lab": "zai",
          "date": "2025-09-30",
          "printed": "14.4",
          "where": "Chart 'LLM Performance Evaluation: Agentic, Reasoning and Coding', panel 'HLE', bar 'GLM-4.5' (base label; only GLM-4.6's separate value is marked 'w/ Tools')"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/9690be94d5b435fb.png"
    },
    {
      "id": "deepseek-v3-1:hle:no-tools-text-only:1",
      "model": "deepseek-v3-1",
      "model_name": "DeepSeek-V3.1",
      "model_date": "2025-08-21",
      "event": "https://frontierlog.ai/event/2025-08-21-deepseek-v3-1",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 15.9,
      "unit": "%",
      "printed": "15.9",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V3.1",
      "source": "https://api-docs.deepseek.com/news/news250922/",
      "source_title": "DeepSeek-V3.1-Terminus",
      "source_date": "2025-09-22",
      "where": "later report: DeepSeek-V3.1-Terminus post, table image, section 'reasoning mode w/o tool use', row \"Humanity's Last Exam\", column 'DeepSeek-V3.1'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "reasoning mode; full or text-only set not stated",
      "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-V3.1",
          "name": "DeepSeek-V3.1 card",
          "title": "deepseek-ai/DeepSeek-V3.1 model card",
          "lab": "deepseek",
          "date": "2025-08-21",
          "printed": "15.9",
          "where": "Evaluation table, General, row 'Humanity's Last Exam (Pass@1)', column 'DeepSeek V3.1-Thinking' (note: 'HLE is evaluated with the text-only subset')"
        },
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.1-Terminus",
          "name": "DeepSeek-V3.1-Terminus card",
          "title": "deepseek-ai/DeepSeek-V3.1-Terminus model card",
          "lab": "deepseek",
          "date": "2025-09-22",
          "printed": "15.9",
          "where": "Table, 'Reasoning Mode w/o Tool Use', row 'Humanity's Last Exam', column 'DeepSeek-V3.1'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/93e083b75de124e9.png",
      "source_image_original": "https://api-docs.deepseek.com/zh-cn/img/deepseek_v3.1_terminus_benchmark_en.png"
    },
    {
      "id": "kimi-k2-thinking:hle:no-tools-text-only:1",
      "model": "kimi-k2-thinking",
      "model_name": "Kimi K2 Thinking",
      "model_date": "2025-11-06",
      "event": "https://frontierlog.ai/event/2025-11-06-kimi-k2-thinking",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 23.9,
      "unit": "%",
      "printed": "23.9",
      "reported_by": "moonshot",
      "run_by": null,
      "label": "K2 Thinking",
      "source": "https://www.kimi.ai/blog/kimi-k2-thinking",
      "source_title": "Introducing Kimi K2 Thinking",
      "source_date": "2025-11-06",
      "where": "'Full Evaluations' table, row 'Humanity's Last Exam (Text-only)', sub-row 'no tools', column 'K2 Thinking'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; 96K thinking-token budget; INT4 precision",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://api-docs.deepseek.com/news/news251201/",
          "name": "DeepSeek-V3.2",
          "title": "DeepSeek-V3.2 Release",
          "lab": "deepseek",
          "date": "2025-12-01",
          "printed": "23.9",
          "where": "Reasoning table image (token counts in parentheses omitted); row 'HLE'; column 'Kimi-K2 Thinking'"
        },
        {
          "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": "23.9",
          "where": "Benchmark table, row 'HLE (separate from 'HLE w/ Tools' row)', column 'Kimi K2 Thinking'"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/4e2626572b0cef3e.png"
    },
    {
      "id": "deepseek-v3-2:hle:no-tools-text-only:1",
      "model": "deepseek-v3-2",
      "model_name": "DeepSeek-V3.2",
      "model_date": "2025-12-01",
      "event": "https://frontierlog.ai/event/2025-12-01-deepseek-v3-2",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 25.1,
      "unit": "%",
      "printed": "25.1",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V3.2-Thinking",
      "source": "https://api-docs.deepseek.com/news/news251201/",
      "source_title": "DeepSeek-V3.2 Release",
      "source_date": "2025-12-01",
      "where": "benchmark chart image (first image), group 'HLE (Pass@1)', bar 'DeepSeek-V3.2-Thinking', value label",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "thinking mode; 21K output tokens; full or text-only set not stated",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-17",
      "check": "human",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "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": "25.1",
          "where": "Score chart image (score.png), panel Humanity's Last Exam, value label on the DeepSeek-V3.2 bar"
        },
        {
          "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": "25.1",
          "where": "Benchmark table, row 'HLE (separate from 'HLE w/ Tools' row)', column 'DeepSeek-V3.2'"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/8ed4c04255cbd56a.webp",
      "source_image_original": "https://api-docs.deepseek.com/img/v3.2_251201_benchmark.webp"
    },
    {
      "id": "deepseek-v3-2-speciale:hle:no-tools-text-only:1",
      "model": "deepseek-v3-2-speciale",
      "model_name": "DeepSeek-V3.2-Speciale",
      "model_date": "2025-12-01",
      "event": "https://frontierlog.ai/event/2025-12-01-deepseek-v3-2",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 30.6,
      "unit": "%",
      "printed": "30.6",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V3.2 Speciale",
      "source": "https://api-docs.deepseek.com/news/news251201/",
      "source_title": "DeepSeek-V3.2 Release",
      "source_date": "2025-12-01",
      "where": "reasoning table image (v3.2_251201_benchmark_table_en), row 'HLE', column 'DeepSeek-V3.2 Speciale', score before the bracketed token count (the first chart image, group 'HLE (Pass@1)', bar 'DeepSeek-V3.2-Speciale', prints the same value label)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "35K output tokens; full or text-only set not stated",
      "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-V3.2/resolve/main/assets/paper.pdf",
          "name": "DeepSeek-V3.2 card",
          "title": "DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models",
          "lab": "deepseek",
          "date": "2025-12-01",
          "printed": "30.6",
          "where": "Table 3, row 'HLE (Pass@1)', column 'DeepSeek-V3.2 Speciale', printed '30.6 (35k)' (also Figure 1)"
        },
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale",
          "name": "DeepSeek-V3.2 card",
          "title": "deepseek-ai/DeepSeek-V3.2-Speciale model card",
          "lab": "deepseek",
          "date": "2025-12-01",
          "printed": "30.6",
          "where": "Benchmark chart (assets/benchmark.png), HLE (Pass@1), DeepSeek-V3.2-Speciale bar label"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/a313f1f47e0fe2c0.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v3.2_251201_benchmark_table_en.png"
    },
    {
      "id": "nemotron-3-nano-30b-a3b:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 10.89,
      "unit": "%",
      "printed": "10.89",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron 3 Nano",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/942b56629bd9b52c6813bb5f68bb5d10ba8cca68/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 model card (revision of 2026-08-11)",
      "source_date": "2026-08-11",
      "where": "'Reasoning Benchmark Evaluations' table, row 'HLE (text-only, no tools)', column 'Nemotron 3 Nano*'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "later re-run; text-only subset",
      "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/5c98520e2f23ada0.png"
    },
    {
      "id": "qwen3-max-thinking:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 30.2,
      "unit": "%",
      "printed": "30.2",
      "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', column 'Qwen3-Max-Thinking'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "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": "glm-5:hle:no-tools-text-only:1",
      "model": "glm-5",
      "model_name": "GLM-5",
      "model_date": "2026-02-12",
      "event": "https://frontierlog.ai/event/2026-02-12-glm-5",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 30.5,
      "unit": "%",
      "printed": "30.5",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5",
      "source": "https://z.ai/blog/glm-5",
      "source_title": "GLM-5: From Vibe Coding to Agentic Engineering",
      "source_date": "2026-02-12",
      "where": "benchmark table, row 'Humanity's Last Exam', column 'GLM-5 (Thinking)'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; GPT-5.2 judge; up to 131,072 output 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/8709485a935545d0.png"
    },
    {
      "id": "nemotron-3-super-120b-a12b:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 20.3,
      "unit": "%",
      "printed": "20.30",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron 3 Super 120B-A12B-BF16",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Bolt 74B-A9B 24GB: A Hybrid Vision-Language Model for Local AI",
      "source_date": "2026-10-15",
      "where": "Table 7 (PDF page 16), row 'HLE (no tools)', column 'Nemotron 3 Super 120B-A12B-BF16' (last column)",
      "own": false,
      "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/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md",
          "name": "Nemotron 3.5 Lightning card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2026-08-11",
          "printed": "20.30",
          "where": "Reasoning Benchmark Evaluations table, HLE (text-only, no tools) row, 'Nemotron 3 Super' column"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/29d0406ec75804d7.png"
    },
    {
      "id": "glm-5-1:hle:no-tools-text-only:1",
      "model": "glm-5-1",
      "model_name": "GLM-5.1",
      "model_date": "2026-04-07",
      "event": "https://frontierlog.ai/event/2026-04-07-glm-5-1",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 31.0,
      "unit": "%",
      "printed": "31.0",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.1",
      "source": "https://z.ai/blog/glm-5.1",
      "source_title": "GLM-5.1: Towards Long-Horizon Tasks",
      "source_date": "2026-04-07",
      "where": "benchmark table near the end, row 'HLE', column 'GLM-5.1'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; GPT-5.2 judge; up to 163,840 output tokens",
      "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": "31.0",
          "where": "Chart 'LLM Performance Evaluation' (all models at max thinking effort), panel 'Humanity's Last Exam', bar 'GLM-5.1' (lower label, not marked w/ Tools)"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/35c66845b36400f8.png"
    },
    {
      "id": "nemotron-3-ultra-550b-a55b:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 29.24,
      "unit": "%",
      "printed": "29.24",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "N3-Ultra 550B-A55B",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI",
      "source_date": "2026-10-15",
      "where": "Table 10 (PDF page 23), row 'HLE (no tools, text)', column 'N3-Ultra 550B-A55B' (second column)",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; GPT-5.6 Luna judge",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16/blob/1f8953f7f0935358e3605150d22bc15df062061c/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-16)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "29.2",
          "where": "evaluation table, 'Nemotron 3 Ultra' column, 'HLE, text-only' row"
        },
        {
          "url": "https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Super-and-Lightning/",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "NVIDIA Nemotron 3.5 Super and Lightning",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "29.2",
          "where": "overview chart, HLE (no tools) group, 'N3 Ultra 550B-A55B' bar label"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/ee443a24cf8af62e.png"
    },
    {
      "id": "glm-5-2:hle:no-tools-text-only:1",
      "model": "glm-5-2",
      "model_name": "GLM-5.2",
      "model_date": "2026-06-16",
      "event": "https://frontierlog.ai/event/2026-06-16-glm-5-2",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 40.5,
      "unit": "%",
      "printed": "40.5",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.2",
      "source": "https://z.ai/blog/glm-5.2",
      "source_title": "GLM-5.2: Built for Long-Horizon Tasks",
      "source_date": "2026-06-16",
      "where": "'Full Benchmark Table', row 'HLE', column 'GLM-5.2'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; GPT-5.5 judge; up to 163,840 output 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/112d3aba81edfc46.png"
    },
    {
      "id": "hy3:hle:no-tools-text-only:1",
      "model": "hy3",
      "model_name": "Hy3",
      "model_date": "2026-07-06",
      "event": "https://frontierlog.ai/event/2026-07-06-hy3",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 37.0,
      "unit": "%",
      "printed": "37.0",
      "reported_by": "tencent",
      "run_by": null,
      "label": "Hy3",
      "source": "https://hy.tencent.ai/research/hy3",
      "source_title": "Introducing Hy3",
      "source_date": "2026-07-06",
      "where": "'Appendix: Benchmark Results' table image, row 'HLE (no tools, text-only)', column 'Hy3'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; highest reasoning setting",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-11-05",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/tencent/Hy3",
          "name": "Hy3 card",
          "title": "tencent/Hy3 model card",
          "lab": "tencent",
          "date": "2026-07-06",
          "printed": "37.0",
          "where": "Model card, Appendix image, row HLE (no tools, text-only), column Hy3. Same table as the Hy3 page's appendix, but a different image file (the page's is d1af768b…). The card's top bar chart has the same bytes as the page's chart, so it is not cited. Mirrored with identical bytes on huggingface.co/tencent/Hy3-FP8 and github.com/Tencent-Hunyuan/Hy3 (assets/benchmark-appendix.png). Notes: highest reasoning tier; SWE-agent scaffold; Terminus-2, 4h timeout, 500 episodes; MCP-Atlas 500-task public set, 100-call budget, Gemini 2.5 Pro judge; BrowseComp self-summary context management."
        }
      ],
      "source_image": "https://frontierlog.ai/sources/d1af768b6734a711.png",
      "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026070622113393_004bbe5190853d7c35825b9257672dde.png?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1783347098;2098707098&q-key-time=1783347098;2098707098&q-header-list=&q-url-param-list=&q-signature=507857321cef2b9fef5e4ef8a8f9ab8937d2480b"
    },
    {
      "id": "hy3:hle:no-tools-text-only:2",
      "model": "hy3",
      "model_name": "Hy3",
      "model_date": "2026-07-06",
      "event": "https://frontierlog.ai/event/2026-07-06-hy3",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 30.0,
      "unit": "%",
      "printed": "30.0",
      "reported_by": "tencent",
      "run_by": null,
      "label": "Hy3 preview",
      "source": "https://hy.tencent.ai/research/hy3",
      "source_title": "Introducing Hy3",
      "source_date": "2026-07-06",
      "where": "'Appendix: Benchmark Results' table image, row 'HLE (no tools, text-only)', column 'Hy3 preview'",
      "own": true,
      "primary": false,
      "parallel": false,
      "notes": "Hy3 preview, re-evaluated at Hy3 launch; text-only subset; highest reasoning setting",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-11-05",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/tencent/Hy3",
          "name": "Hy3 card",
          "title": "tencent/Hy3 model card",
          "lab": "tencent",
          "date": "2026-07-06",
          "printed": "30.0",
          "where": "Model card, Appendix image, row HLE (no tools, text-only), column Hy3 preview. Same table as the Hy3 page's appendix, but a different image file (the page's is d1af768b…). The card's top bar chart has the same bytes as the page's chart, so it is not cited. Mirrored with identical bytes on huggingface.co/tencent/Hy3-FP8 and github.com/Tencent-Hunyuan/Hy3 (assets/benchmark-appendix.png). Notes: highest reasoning tier; SWE-agent scaffold; Terminus-2, 4h timeout, 500 episodes; MCP-Atlas 500-task public set, 100-call budget, Gemini 2.5 Pro judge; BrowseComp self-summary context management."
        },
        {
          "url": "https://hy.tencent.ai/research/hy3-preview",
          "name": "Tencent releases Hy3 preview",
          "title": "Hy3 preview: The First Step in Rebuilding the Hy model",
          "lab": "tencent",
          "date": "2026-04-23",
          "printed": "30.0",
          "where": "Appendix table image, row HLE (setting: no tools, text-only), column Hy3 preview. Also the page's STEM bar chart, HLE panel, Hy3 preview 30.0 (sha256 1f70fbce5d9945de7b9eb54d0560b1f1bb66f3a067493238b680506c53b8bcb3). Hy3 preview launch page (Apr 23, 2026), appendix table, 'Hy3 preview' column. Notes: competitors at high-think; 'For the SWE-Bench series ... SWE-agent scaffold, with the best results obtained in non-thinking mode.'"
        },
        {
          "url": "https://huggingface.co/tencent/Hy3-preview",
          "name": "Tencent releases Hy3 preview card",
          "title": "tencent/Hy3-preview model card",
          "lab": "tencent",
          "date": "2026-04-23",
          "printed": "30.0",
          "where": "Hy3-preview model card, image bench_stem.jpg, HLE panel, Hy3 preview bar 30.0. The caption says domestic models were run on the text-only subset."
        }
      ],
      "source_image": "https://frontierlog.ai/sources/d1af768b6734a711.png",
      "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026070622113393_004bbe5190853d7c35825b9257672dde.png?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1783347098;2098707098&q-key-time=1783347098;2098707098&q-header-list=&q-url-param-list=&q-signature=507857321cef2b9fef5e4ef8a8f9ab8937d2480b"
    },
    {
      "id": "hy3:hle:no-tools-text-only:3",
      "model": "hy3",
      "model_name": "Hy3",
      "model_date": "2026-07-06",
      "event": "https://frontierlog.ai/event/2026-07-06-hy3",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 34.4,
      "unit": "%",
      "printed": "34.4",
      "reported_by": "tencent",
      "run_by": null,
      "label": "Hy3",
      "source": "https://hy.tencent.ai/research/hy4-preview",
      "source_title": "Introducing Hy4 preview",
      "source_date": "2026-08-28",
      "where": "later report: Hy4 preview post, 'Appendix: Benchmark Results' table image, row 'HLE (no tools, text-only)', column 'Hy3'",
      "own": false,
      "primary": false,
      "parallel": false,
      "notes": "re-evaluated at Hy4 preview launch; text-only subset; highest reasoning setting",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-11-07",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/02c751bf51c52c52.jpg",
      "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026082812342438_ecb98509e92e4be02a40ca358414133a.jpg?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1787891670;2103251670&q-key-time=1787891670;2103251670&q-header-list=&q-url-param-list=&q-signature=b6d0c3c4d6d7e0e5219d177ec730735185ccdd1f"
    },
    {
      "id": "nemotron-3-5-lightning-30b-a3b:hle:no-tools-text-only:1",
      "model": "nemotron-3-5-lightning-30b-a3b",
      "model_name": "Nemotron 3.5 Lightning 30B-A3B",
      "model_date": "2026-08-11",
      "event": "https://frontierlog.ai/event/2026-08-11-nemotron-3-5-lightning",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 11.72,
      "unit": "%",
      "printed": "11.72",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron-3.5-Lightning-30B-A3B-BF16",
      "source": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4/blob/b14872a58afd94e2fbbc1e5a1fe4ebb7dfdc5fdd/README.md",
      "source_title": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4 model card (launch-day revision)",
      "source_date": "2026-08-11",
      "where": "'Reasoning Benchmark Evaluations' table, row 'HLE (text-only, no tools)', column 'Nemotron-3.5-Lightning-30B-A3B-BF16'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "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/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16/blob/a9904d24bcc1d289a1950fa9d2b978c47cf903b9/README.md",
          "name": "Nemotron 3.5 Lightning card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 model card (revision of 2026-08-24)",
          "lab": "nvidia",
          "date": "2026-08-11",
          "printed": "11.72",
          "where": "Reasoning Benchmark Evaluations table, HLE (text-only, no tools) row, BF16 column"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/662f596717868384.png"
    },
    {
      "id": "nemotron-3-5-lightning-30b-a3b:hle:no-tools-text-only:2",
      "model": "nemotron-3-5-lightning-30b-a3b",
      "model_name": "Nemotron 3.5 Lightning 30B-A3B",
      "model_date": "2026-08-11",
      "event": "https://frontierlog.ai/event/2026-08-11-nemotron-3-5-lightning",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 10.8,
      "unit": "%",
      "printed": "10.80",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Lightning 3.5 BF16",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI",
      "source_date": "2026-10-15",
      "where": "Table 22 (PDF page 42), row 'HLE (no tools)', column 'BF16 acc.'",
      "own": false,
      "primary": false,
      "parallel": false,
      "notes": "average of 2 runs",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://developer.nvidia.com/blog/developing-nemotron-3-5-lightning-nvfp4-with-qad-using-nvidia-model-optimizer/",
          "name": "Nemotron 3.5 Lightning card",
          "title": "Developing Nemotron 3.5 Lightning NVFP4 with QAD Using NVIDIA Model Optimizer",
          "lab": "nvidia",
          "date": "2026-08-11",
          "printed": "10.80",
          "where": "Table 4, HLE (w/o tools) row, BF16 column"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/0916f77b2b4a2e75.png"
    },
    {
      "id": "deepseek-v4-pro-0813:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 42.7,
      "unit": "%",
      "printed": "42.7",
      "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', first 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": "42.7",
          "where": "Benchmark table image, row 'HLE', column 'DeepSeek V4-Pro 0813', printed '42.7*' (* = text-only subset of HLE)"
        },
        {
          "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": "42.7",
          "where": "'Comparison with frontier models (Max reasoning effort)' table, row 'HLE (Pass@1)', column 'DS-V4-Pro', printed '42.7†' († = text-only subset)"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/51578c812ffc7347.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260813_benchmark_table_en.png"
    },
    {
      "id": "hy4-preview:hle:no-tools-text-only:1",
      "model": "hy4-preview",
      "model_name": "Hy4 preview",
      "model_date": "2026-08-28",
      "event": "https://frontierlog.ai/event/2026-08-28-hy4-preview",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 43.4,
      "unit": "%",
      "printed": "43.4",
      "reported_by": "tencent",
      "run_by": null,
      "label": "Hy4 preview",
      "source": "https://hy.tencent.ai/research/hy4-preview",
      "source_title": "Introducing Hy4 preview",
      "source_date": "2026-08-28",
      "where": "'Appendix: Benchmark Results' table image, row 'HLE (no tools, text-only)', column 'Hy4 preview'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; highest reasoning setting",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-11-09",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/tencent/Hy4-preview",
          "name": "Hy4 preview card",
          "title": "tencent/Hy4-preview model card",
          "lab": "tencent",
          "date": "2026-08-28",
          "printed": "43.4",
          "where": "'A New Flagship Generation' bar chart (assets/benchmark.jpg), panel 'Humanity's Last Exam (text-only, no tools)', Hy4 preview (dark blue) bar"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/02c751bf51c52c52.jpg",
      "source_image_original": "https://hy-model-ap-prod-1258344703.cos.accelerate.myqcloud.com/llm-blog/default/9a5ea02f/2026082812342438_ecb98509e92e4be02a40ca358414133a.jpg?q-sign-algorithm=sha1&q-ak=AKIDRl074nOsGdJ9zjMsCRWP3ShmgS3VtX4S&q-sign-time=1787891670;2103251670&q-key-time=1787891670;2103251670&q-header-list=&q-url-param-list=&q-signature=b6d0c3c4d6d7e0e5219d177ec730735185ccdd1f"
    },
    {
      "id": "deepseek-v4-1-flash:hle:no-tools-text-only: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": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 39.1,
      "unit": "%",
      "printed": "39.1",
      "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', column 'DeepSeek V4.1-Flash', value inside the brackets (asterisk)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "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": "39.1",
          "where": "'Comparison with frontier models (Max reasoning effort)' table, row 'HLE (Pass@1)', column 'DS-V4.1-Flash', printed '36.8 (39.1†)'; † = text-only subset"
        },
        {
          "url": "https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017",
          "name": "DeepSeek-V4.1-Flash-1017 card",
          "title": "deepseek-ai/DeepSeek-V4.1-Flash-1017 model card",
          "lab": "deepseek",
          "date": "2026-10-17",
          "printed": "39.1",
          "where": "Table, row 'HLE w/o tools', column 'DS-V4.1-Flash', printed '36.8 (39.1†)'; † = text-only subset"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/19d3b83b1b9e5f5b.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4.1_260910_benchmark_table_en.png"
    },
    {
      "id": "nemotron-3-5-bolt:hle:no-tools-text-only:1",
      "model": "nemotron-3-5-bolt",
      "model_name": "Nemotron 3.5 Bolt",
      "model_date": "2026-10-15",
      "event": "https://frontierlog.ai/event/2026-10-15-nemotron-3-5-super-vl-bolt",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 28.73,
      "unit": "%",
      "printed": "28.73",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron 3.5 Bolt 74B-A9B-BF16",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Bolt 74B-A9B 24GB: A Hybrid Vision-Language Model for Local AI",
      "source_date": "2026-10-15",
      "where": "Table 7 (section 8), row 'HLE (no tools)', column 'Nemotron 3.5 Bolt 74B-A9B-BF16'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "BF16 weights",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-26",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16/blob/567021037ff88364143e24ea05e6abd0f72086ff/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16 model card (revision of 2026-10-19)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "28.73",
          "where": "BF16 HF card Benchmarks table, row \"HLE (no tools)\", column \"Nemotron 3.5 Bolt 74B-A9B (BF16)\" (same value in launch revision ea56833 of 2026-10-15)."
        }
      ]
    },
    {
      "id": "nemotron-3-5-bolt:hle:no-tools-text-only:2",
      "model": "nemotron-3-5-bolt",
      "model_name": "Nemotron 3.5 Bolt",
      "model_date": "2026-10-15",
      "event": "https://frontierlog.ai/event/2026-10-15-nemotron-3-5-super-vl-bolt",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 25.3,
      "unit": "%",
      "printed": "25.30",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron 3.5 Bolt 74B-A9B-24GB",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Bolt 74B-A9B 24GB: A Hybrid Vision-Language Model for Local AI",
      "source_date": "2026-10-15",
      "where": "Table 7 (section 8), row 'HLE (no tools)', column 'Nemotron 3.5 Bolt 74B-A9B-24GB'",
      "own": true,
      "primary": false,
      "parallel": false,
      "notes": "24GB build with 2-bit (IQ2) weights",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-26",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16/blob/567021037ff88364143e24ea05e6abd0f72086ff/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16 model card (revision of 2026-10-19)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "25.30",
          "where": "BF16 HF card Benchmarks table, row \"HLE (no tools)\", column \"Nemotron 3.5 Bolt 74B-A9B-24GB (IQ2)\"."
        }
      ]
    },
    {
      "id": "nemotron-3-5-super-vl-120b-a12b:hle:no-tools-text-only:1",
      "model": "nemotron-3-5-super-vl-120b-a12b",
      "model_name": "Nemotron 3.5 Super VL 120B-A12B",
      "model_date": "2026-10-15",
      "event": "https://frontierlog.ai/event/2026-10-15-nemotron-3-5-super-vl-bolt",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 31.79,
      "unit": "%",
      "printed": "31.79",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "N3.5-Super 120B-A12B",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Super-and-Lightning-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI",
      "source_date": "2026-10-15",
      "where": "Table 10 (PDF page 23), row 'HLE (no tools, text)', column 'N3.5-Super 120B-A12B' (first column)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset; GPT-5.6 Luna judge",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-19",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16/blob/1f8953f7f0935358e3605150d22bc15df062061c/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 model card (revision of 2026-10-16)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "31.79",
          "where": "evaluation table, 'Nemotron 3.5 SuperGA BF16' column, 'HLE, text-only' row"
        },
        {
          "url": "https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4/blob/59b8721646c38111a242da18218859fb2b454322/README.md",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4 model card (revision of 2026-10-16)",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "31.79",
          "where": "BF16 vs NVFP4 table, 'HLE, text-only' row, 'Nemotron 3.5 Super VL BF16' column"
        },
        {
          "url": "https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Super-and-Lightning/",
          "name": "Nemotron 3.5 Super VL and Bolt card",
          "title": "NVIDIA Nemotron 3.5 Super and Lightning",
          "lab": "nvidia",
          "date": "2026-10-15",
          "printed": "31.8",
          "where": "overview chart, HLE (no tools) group, 'N3.5 Super 120B-A12B' bar label"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/ee443a24cf8af62e.png"
    },
    {
      "id": "nemotron-3-5-super-vl-120b-a12b:hle:no-tools-text-only:2",
      "model": "nemotron-3-5-super-vl-120b-a12b",
      "model_name": "Nemotron 3.5 Super VL 120B-A12B",
      "model_date": "2026-10-15",
      "event": "https://frontierlog.ai/event/2026-10-15-nemotron-3-5-super-vl-bolt",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 31.84,
      "unit": "%",
      "printed": "31.84",
      "reported_by": "nvidia",
      "run_by": null,
      "label": "Nemotron 3.5 Super 120B-A12B-BF16",
      "source": "https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB-Technical-Report.pdf",
      "source_title": "Nemotron 3.5 Bolt 74B-A9B 24GB: A Hybrid Vision-Language Model for Local AI",
      "source_date": "2026-10-15",
      "where": "Table 7 (PDF page 16), row 'HLE (no tools)', column 'Nemotron 3.5 Super 120B-A12B-BF16' (first column)",
      "own": true,
      "primary": false,
      "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": "31.84",
          "where": "Table 16 (PTQ/QAD at 4.8 BPE), HLE w/o tools row, BF16 column; also Table 13 BF16 column"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/29d0406ec75804d7.png"
    },
    {
      "id": "deepseek-v4-1-flash-1017:hle:no-tools-text-only:1",
      "model": "deepseek-v4-1-flash-1017",
      "model_name": "DeepSeek-V4.1-Flash-1017",
      "model_date": "2026-10-17",
      "event": "https://frontierlog.ai/event/2026-10-17-deepseek-v4-1-flash-1017",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 44.0,
      "unit": "%",
      "printed": "44.0",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek V4.1-Flash-1017",
      "source": "https://api-docs.deepseek.com/news/news261017/",
      "source_title": "DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made Real",
      "source_date": "2026-10-17",
      "where": "benchmark table image (first image), row 'HLE (w/o tools)', column 'DeepSeek V4.1-Flash-1017', second of the two values (asterisk)",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "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-1017",
          "name": "DeepSeek-V4.1-Flash-1017 card",
          "title": "deepseek-ai/DeepSeek-V4.1-Flash-1017 model card",
          "lab": "deepseek",
          "date": "2026-10-17",
          "printed": "44.0",
          "where": "Table, row 'HLE w/o tools', column 'DS-V4.1-Flash-1017', printed '42.6 (44.0†)'; † = text-only subset"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/52c22094008778b4.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4.1_261017_benchmark_table_en.png"
    },
    {
      "id": "minimax-m3-1-flash:hle:no-tools-text-only:1",
      "model": "minimax-m3-1-flash",
      "model_name": "MiniMax M3.1 Flash",
      "model_date": "2026-10-19",
      "event": "https://frontierlog.ai/event/2026-10-19-minimax-m3-1-flash",
      "benchmark": "hle",
      "benchmark_name": "Humanity's Last Exam",
      "setting": "no-tools-text-only",
      "setting_name": "No tools, text-only subset",
      "score": 43.4,
      "unit": "%",
      "printed": "43.4",
      "reported_by": "minimax",
      "run_by": null,
      "label": "MiniMax M3.1 Flash",
      "source": "https://www.minimax.io/blog/minimax-m31f",
      "source_title": "MiniMax M3.1 Flash: Redefining Agent Performance and Cost",
      "source_date": "2026-10-19",
      "where": "benchmark table image (first image), row 'HLE (w/o tools)', column 'MiniMax M3.1 Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "text-only subset",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-26",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/5525d0bcceba0858.m31-flash-en-2-1792403273121-webp",
      "source_image_original": "https://filecdn.minimax.chat/public/m31-flash-en-2-1792403273121-webp-1.m31-flash-en-2-1792403273121-webp"
    }
  ]
}