{
  "api_version": "1",
  "updated": "2026-10-29",
  "latest_event": "2026-10-28",
  "rights": "© Frontier Log, all rights reserved.",
  "benchmark": {
    "slug": "agents-last-exam",
    "name": "Agents' Last Exam v1",
    "family": "agents-last-exam",
    "unit": "%",
    "higher_is_better": true,
    "status": "superseded",
    "active": false,
    "superseded_by": "agents-last-exam-1-1",
    "retired": null,
    "benchmark_url": "https://agents-last-exam.org",
    "introduced": "2026-06-03",
    "creator": "UC Berkeley RDI",
    "introduced_source": "https://arxiv.org/abs/2606.05405v1",
    "settings": [
      {
        "setting": "pass",
        "name": "Pass rate",
        "url": "https://frontierlog.ai/benchmark/agents-last-exam",
        "results": 8
      },
      {
        "setting": "score",
        "name": "Score",
        "url": "https://frontierlog.ai/benchmark/agents-last-exam/score",
        "results": 4
      },
      {
        "setting": "cli-pass",
        "name": "CLI subset, pass rate",
        "url": "https://frontierlog.ai/benchmark/agents-last-exam/cli-pass",
        "results": 8
      },
      {
        "setting": "unstated",
        "name": "Measure not stated",
        "url": "https://frontierlog.ai/benchmark/agents-last-exam/unstated",
        "results": 8
      }
    ]
  },
  "total": 28,
  "results": [
    {
      "id": "gpt-5-5:agents-last-exam: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 46.9,
      "unit": "%",
      "printed": "46.9%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.5",
      "source": "https://openai.com/index/gpt-5-6",
      "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "source_date": "2026-07-09",
      "where": "later report: GPT-5.6 post, professional eval table (first table), row 'Agents' Last Exam', column 'GPT-5.5'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "effort and harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "deepseek-v4-flash:agents-last-exam:cli-pass:1",
      "model": "deepseek-v4-flash",
      "model_name": "DeepSeek-V4-Flash",
      "model_date": "2026-04-24",
      "event": "https://frontierlog.ai/event/2026-04-24-deepseek-v4-preview",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 15.8,
      "unit": "%",
      "printed": "15.8",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V4-Flash-Preview",
      "source": "https://api-docs.deepseek.com/news/news260813/",
      "source_title": "DeepSeek-V4-Pro GA Release",
      "source_date": "2026-08-13",
      "where": "later report: DeepSeek-V4-Pro-0813 post, table image, row 'Agents' Last Exam', column 'DeepSeek-V4-Flash-Preview'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/51578c812ffc7347.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260813_benchmark_table_en.png"
    },
    {
      "id": "deepseek-v4-pro:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 16.5,
      "unit": "%",
      "printed": "16.5",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek-V4-Pro-Preview",
      "source": "https://api-docs.deepseek.com/news/news260813/",
      "source_title": "DeepSeek-V4-Pro GA Release",
      "source_date": "2026-08-13",
      "where": "later report: DeepSeek-V4-Pro-0813 post, table image, row 'Agents' Last Exam', column 'DeepSeek-V4-Pro-Preview'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/51578c812ffc7347.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260813_benchmark_table_en.png"
    },
    {
      "id": "qwen3-7-max:agents-last-exam:pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 11.8,
      "unit": "%",
      "printed": "11.8",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.7-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": "later report: Qwen3.8-Max post, 'Full Benchmark Table' (first table), General Agent block, row 'Agents' Last Exam (Pass / Score)', column 'Qwen3.7-Max', first of the two values",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "qwen3-7-max:agents-last-exam:score: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "score",
      "setting_name": "Score",
      "score": 31.1,
      "unit": "%",
      "printed": "31.1",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.7-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": "later report: Qwen3.8-Max post, 'Full Benchmark Table' (first table), General Agent block, row 'Agents' Last Exam (Pass / Score)', column 'Qwen3.7-Max', second of the two values",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "qwen3-7-plus:agents-last-exam:pass:1",
      "model": "qwen3-7-plus",
      "model_name": "Qwen3.7-Plus",
      "model_date": "2026-06-01",
      "event": "https://frontierlog.ai/event/2026-06-01-qwen3-7-plus",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 13.2,
      "unit": "%",
      "printed": "13.2",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.7-Plus",
      "source": "https://qwen.ai/blog?id=qwen3.8-flash-next",
      "source_title": "Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency",
      "source_date": "2026-08-26",
      "where": "later report: Qwen3.8-Flash-Next post, Language table (first table), Agent block, row 'Agents' Last Exam', column 'Qwen3.7-Plus', the value labelled 'Pass@1'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "pass@1; harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
          "name": "Qwen3.8-Flash-Next card",
          "title": "Qwen/Qwen3.8-Flash-Next model card",
          "lab": "alibaba",
          "date": "2026-08-26",
          "printed": "13.2",
          "where": "model card benchmark table, row \"Frontier agentic tasks / Agents' Last Exam\", column \"Qwen3.7-Plus\", sub-value labelled \"Pass@1\""
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/1ce175beb0b20cf1.png"
    },
    {
      "id": "qwen3-7-plus:agents-last-exam:score:1",
      "model": "qwen3-7-plus",
      "model_name": "Qwen3.7-Plus",
      "model_date": "2026-06-01",
      "event": "https://frontierlog.ai/event/2026-06-01-qwen3-7-plus",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "score",
      "setting_name": "Score",
      "score": 33.6,
      "unit": "%",
      "printed": "33.6",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.7-Plus",
      "source": "https://qwen.ai/blog?id=qwen3.8-flash-next",
      "source_title": "Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency",
      "source_date": "2026-08-26",
      "where": "later report: Qwen3.8-Flash-Next post, Language table (first table), Agent block, row 'Agents' Last Exam', column 'Qwen3.7-Plus', the value labelled 'Score'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
          "name": "Qwen3.8-Flash-Next card",
          "title": "Qwen/Qwen3.8-Flash-Next model card",
          "lab": "alibaba",
          "date": "2026-08-26",
          "printed": "33.6",
          "where": "model card benchmark table, row \"Frontier agentic tasks / Agents' Last Exam\", column \"Qwen3.7-Plus\", sub-value labelled \"Score\""
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/1ce175beb0b20cf1.png"
    },
    {
      "id": "glm-5-2:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 23.8,
      "unit": "%",
      "printed": "23.8",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.2",
      "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": "later report: GLM-5.3 post, 'Performance across comparison models' table, row 'Agents' Last Exam' / 'ALE-CLI', column 'GLM-5.2'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "105-task CLI subset; Claude Code harness",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "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": "23.8",
          "where": "Benchmark table, row 'Agents' Last Exam (ALE-CLI)', column 'GLM-5.2'; footnote 'Agent's Last Exam (CLI)': Claude Code harness, 105 tasks"
        },
        {
          "url": "https://github.com/zai-org/GLM-5",
          "name": "GLM-5 card",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering (zai-org/GLM-5 GitHub repository)",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "23.8",
          "where": "README 'GLM-5.3 & GLM-5.3-Flash' section, image resources/bench_53_2.png, Agents' Last Exam panel, GLM-5.2 bar"
        }
      ]
    },
    {
      "id": "glm-5-2:agents-last-exam:unstated: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 20.4,
      "unit": "%",
      "printed": "20.4",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.2",
      "source": "https://z.ai/blog/glm-5.3-flash",
      "source_title": "GLM-5.3-Flash: Frontier Intelligence, Flash Cost",
      "source_date": "2026-08-26",
      "where": "later report: GLM-5.3-Flash post, main benchmark table (second table), AGENTIC block, row 'Agents' Last Exam', column 'GLM-5.2'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "Claude Code harness; subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/zai-org/GLM-5.3-Flash",
          "name": "GLM-5.3-Flash card",
          "title": "zai-org/GLM-5.3-Flash model card",
          "lab": "zai",
          "date": "2026-08-26",
          "printed": "20.4",
          "where": "bench_53.png chart 'LLM Performance Evaluation' (subtitle 'Agents' Last Exam', no subset), panel 'Agents' Last Exam', GLM-5.2 bar"
        },
        {
          "url": "https://github.com/zai-org/GLM-5",
          "name": "GLM-5 card",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering (zai-org/GLM-5 GitHub repository)",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "20.4",
          "where": "README, image resources/bench_53.png (GLM-5.3-Flash chart), Agents' Last Exam panel, GLM-5.2 bar"
        },
        {
          "url": "https://docs.z.ai/guides/llm/glm-5.3-flash",
          "name": "GLM-5.3-Flash card",
          "title": "GLM-5.3-Flash/FlashX - Overview",
          "lab": "zai",
          "date": "2026-08-26",
          "printed": "20.4",
          "where": "Benchmark image (GLM-5.3-Flash chart), panel 'Agents' Last Exam', GLM-5.2 bar"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/58a995e241d3e40c.png"
    },
    {
      "id": "gpt-5-6-luna:agents-last-exam:unstated:1",
      "model": "gpt-5-6-luna",
      "model_name": "GPT-5.6 Luna",
      "model_date": "2026-07-09",
      "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 50.3,
      "unit": "%",
      "printed": "50.3%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.6 Luna",
      "source": "https://openai.com/index/gpt-5-6",
      "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "source_date": "2026-07-09",
      "where": "professional eval table (first table), row 'Agents' Last Exam', column 'GPT-5.6 Luna'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "effort and harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "gpt-5-6-sol:agents-last-exam:unstated:1",
      "model": "gpt-5-6-sol",
      "model_name": "GPT-5.6 Sol",
      "model_date": "2026-07-09",
      "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 53.6,
      "unit": "%",
      "printed": "53.6",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT-5.6 Sol",
      "source": "https://openai.com/index/gpt-5-6",
      "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "source_date": "2026-07-09",
      "where": "opening section, paragraph beginning 'We trained GPT-5.6 to get more useful work from every token', the GPT-5.6 Sol figure on Agents' Last Exam",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "headline figure; effort and harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://qwen.ai/blog?id=qwen3.8",
          "name": "Qwen3.8-Max",
          "title": "Qwen3.8-Max: A New Bar for Coding and Cowork",
          "lab": "alibaba",
          "date": "2026-08-03",
          "printed": "53.6",
          "where": "Alibaba (another lab's table): performance chart, panel \"Agents' Last Exam (Score)\", OpenAI-logo bar (legend \"GPT5.6 Sol (max)\")"
        },
        {
          "url": "https://openai.com/index/gpt-6-astra",
          "name": "GPT-6 Astra",
          "title": "GPT-6 Astra: A new generation of intelligence",
          "lab": "openai",
          "date": "2026-09-03",
          "printed": "53.6%",
          "where": "Computer Use eval table (HTML), row 'Agents' Last Exam', column 'GPT‑5.6 Sol' (header carries footnote 2); also the sentence '...compared with 55.5% for Claude Opus 5 and 53.6% for GPT‑5.6 Sol' (highest-scoring settings)"
        }
      ]
    },
    {
      "id": "gpt-5-6-sol:agents-last-exam:unstated:2",
      "model": "gpt-5-6-sol",
      "model_name": "GPT-5.6 Sol",
      "model_date": "2026-07-09",
      "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 52.7,
      "unit": "%",
      "printed": "52.7%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.6 Sol",
      "source": "https://openai.com/index/gpt-5-6",
      "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "source_date": "2026-07-09",
      "where": "professional eval table (first table, columns GPT-5.6 Sol, Terra, Luna, GPT-5.5, ...), row 'Agents' Last Exam', column 'GPT-5.6 Sol'",
      "own": true,
      "primary": false,
      "parallel": false,
      "notes": "eval table figure; effort and harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "gpt-5-6-terra:agents-last-exam:unstated:1",
      "model": "gpt-5-6-terra",
      "model_name": "GPT-5.6 Terra",
      "model_date": "2026-07-09",
      "event": "https://frontierlog.ai/event/2026-07-09-gpt-5-6",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 50.4,
      "unit": "%",
      "printed": "50.4%",
      "reported_by": "openai",
      "run_by": null,
      "label": "GPT‑5.6 Terra",
      "source": "https://openai.com/index/gpt-5-6",
      "source_title": "GPT-5.6: Frontier intelligence that scales with your ambition",
      "source_date": "2026-07-09",
      "where": "professional eval table (first table), row 'Agents' Last Exam', column 'GPT-5.6 Terra'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "effort and harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "gemini-3-6-flash:agents-last-exam:pass:1",
      "model": "gemini-3-6-flash",
      "model_name": "Gemini 3.6 Flash",
      "model_date": "2026-07-21",
      "event": "https://frontierlog.ai/event/2026-07-21-gemini-3-6-flash-and-3-5-flash",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 24.2,
      "unit": "%",
      "printed": "24.2%",
      "reported_by": "google",
      "run_by": null,
      "label": "Gemini 3.6 Flash",
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
      "source_title": "Introducing Gemini 3.7 Flash",
      "source_date": "2026-08-13",
      "where": "later report: Gemini 3.7 Flash post, benchmark table (image), row 'Agent's Last Exam', 'Pass rate', column 'Gemini 3.6 Flash'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "ALE-Claw harness",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://deepmind.google/models/model-cards/gemini-3-7-flash/",
          "name": "Gemini 3.7 Flash card",
          "title": "Gemini 3.7 Flash model card",
          "lab": "google",
          "date": "2026-08-13",
          "printed": "24.2%",
          "where": "Results table (HTML text), row \"Agent's Last Exam / Multimodal desktop and OS agent tasks\", note \"Pass rate\", column Gemini 3.6 Flash"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/73ca55228187a14c.png",
      "source_image_original": "https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini-3-7-flash__evals__benchmarks-table__2x.original.png"
    },
    {
      "id": "qwen3-8-max:agents-last-exam:pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 27.0,
      "unit": "%",
      "printed": "27.0",
      "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 table), General Agent block, row 'Agents' Last Exam (Pass / Score)', column 'Qwen3.8-Max' (last column), first of the two values",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "qwen3-8-max:agents-last-exam:score: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "score",
      "setting_name": "Score",
      "score": 52.4,
      "unit": "%",
      "printed": "52.4",
      "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 table), General Agent block, row 'Agents' Last Exam (Pass / Score)', column 'Qwen3.8-Max' (last column), second of the two values",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "deepseek-v4-pro-0813:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 25.7,
      "unit": "%",
      "printed": "25.7",
      "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 'Agents' Last Exam', column 'DeepSeek-V4-Pro-0813'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "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": "25.7",
          "where": "benchmark table image, row 'Agents' Last Exam', 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": "25.7",
          "where": "'Comparison with frontier models (Max reasoning effort)' table, row 'Agent's Last Exam (Pass@1)', column 'DS-V4-Pro' (unsuffixed label; values match V4-Pro-0813, as the 260910 post's same table labels it 'V4-Pro 0813')"
        },
        {
          "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": "25.7",
          "where": "Z.ai GLM-5.3 card benchmark table, row 'Agents' Last Exam (ALE-CLI)', column 'DeepSeek-V4 Pro-0813' (same table in zai-org/GLM-5.3-BF16)"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/51578c812ffc7347.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260813_benchmark_table_en.png"
    },
    {
      "id": "gemini-3-7-flash:agents-last-exam:pass:1",
      "model": "gemini-3-7-flash",
      "model_name": "Gemini 3.7 Flash",
      "model_date": "2026-08-13",
      "event": "https://frontierlog.ai/event/2026-08-13-gemini-3-7-flash",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 26.3,
      "unit": "%",
      "printed": "26.3%",
      "reported_by": "google",
      "run_by": null,
      "label": "Gemini 3.7 Flash",
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/",
      "source_title": "Introducing Gemini 3.7 Flash",
      "source_date": "2026-08-13",
      "where": "benchmark table (image), row 'Agent's Last Exam', 'Pass rate', column 'Gemini 3.7 Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "ALE-Claw harness",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://deepmind.google/models/model-cards/gemini-3-7-flash/",
          "name": "Gemini 3.7 Flash card",
          "title": "Gemini 3.7 Flash model card",
          "lab": "google",
          "date": "2026-08-13",
          "printed": "26.3%",
          "where": "Results table (HTML text), row \"Agent's Last Exam\", note \"Pass rate\", column Gemini 3.7 Flash"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/73ca55228187a14c.png",
      "source_image_original": "https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini-3-7-flash__evals__benchmarks-table__2x.original.png"
    },
    {
      "id": "glm-5-3:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 28.5,
      "unit": "%",
      "printed": "28.5",
      "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 'Agents' Last Exam' / 'ALE-CLI', column 'GLM-5.3'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "105-task CLI subset; Claude Code harness; max effort; 1M context; 4h timeout, up to 8h per task",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "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": "28.5",
          "where": "Benchmark table, row 'Agents' Last Exam (ALE-CLI)', column 'GLM-5.3'; footnote: Claude Code, effort max, 1M context, 105 tasks, 4h default timeout up to 8h"
        },
        {
          "url": "https://github.com/zai-org/GLM-5",
          "name": "GLM-5 card",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering (zai-org/GLM-5 GitHub repository)",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "28.5",
          "where": "README, image resources/bench_53_2.png, Agents' Last Exam panel, GLM-5.3 bar"
        },
        {
          "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": "28.5",
          "where": "DeepSeek card, table row 'Agents’ Last Exam', column 'GLM-5.3' (subset not stated)"
        },
        {
          "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": "28.5",
          "where": "DeepSeek card, table row 'Agent's Last Exam (Pass@1)', column 'GLM-5.3' (subset not stated)"
        }
      ]
    },
    {
      "id": "deepseek-v4-flash-vision-exp:agents-last-exam:cli-pass:1",
      "model": "deepseek-v4-flash-vision-exp",
      "model_name": "DeepSeek-V4-Flash-Vision-Exp",
      "model_date": "2026-08-21",
      "event": "https://frontierlog.ai/event/2026-08-21-deepseek-v4-flash-vision-exp",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 27.3,
      "unit": "%",
      "printed": "27.3",
      "reported_by": "deepseek",
      "run_by": null,
      "label": "DeepSeek V4-Flash-Vision-Exp",
      "source": "https://api-docs.deepseek.com/news/news260821/",
      "source_title": "DeepSeek-V4-Flash-Vision-Exp Release: Multimodal API Now Live",
      "source_date": "2026-08-21",
      "where": "table image, section 'Multimodal Agent Evaluation', row 'Agents' Last Exam', column 'DeepSeek V4-Flash-Vision-Exp'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/zai-org/GLM-5.3-Flash",
          "name": "GLM-5.3-Flash card",
          "title": "zai-org/GLM-5.3-Flash model card",
          "lab": "zai",
          "date": "2026-08-26",
          "printed": "27.3",
          "where": "Z.ai GLM-5.3-Flash card, 'LLM Performance Evaluation' chart image, panel 'Agents' Last Exam', bar 'DeepSeek-V4-Vision-Exp' (legend label; whale icon)"
        }
      ],
      "source_image": "https://frontierlog.ai/sources/4b3cca9a4075fb05.png",
      "source_image_original": "https://api-docs.deepseek.com/img/v4_260821_benchmark_en.png"
    },
    {
      "id": "glm-5-3-flash:agents-last-exam:unstated:1",
      "model": "glm-5-3-flash",
      "model_name": "GLM-5.3-Flash",
      "model_date": "2026-08-26",
      "event": "https://frontierlog.ai/event/2026-08-26-glm-5-3-flash",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 26.3,
      "unit": "%",
      "printed": "26.3",
      "reported_by": "zai",
      "run_by": null,
      "label": "GLM-5.3-Flash",
      "source": "https://z.ai/blog/glm-5.3-flash",
      "source_title": "GLM-5.3-Flash: Frontier Intelligence, Flash Cost",
      "source_date": "2026-08-26",
      "where": "main benchmark table (second table), AGENTIC block, row 'Agents' Last Exam', column 'GLM-5.3-Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "Claude Code harness; max effort; 1M context; subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/zai-org/GLM-5.3-Flash",
          "name": "GLM-5.3-Flash card",
          "title": "zai-org/GLM-5.3-Flash model card",
          "lab": "zai",
          "date": "2026-08-26",
          "printed": "26.3",
          "where": "bench_53.png chart 'LLM Performance Evaluation' (subtitle 'Agents' Last Exam', no subset), panel 'Agents' Last Exam', GLM-5.3-Flash bar"
        },
        {
          "url": "https://github.com/zai-org/GLM-5",
          "name": "GLM-5 card",
          "title": "GLM-5: From Vibe Coding to Agentic Engineering (zai-org/GLM-5 GitHub repository)",
          "lab": "zai",
          "date": "2026-02-12",
          "printed": "26.3",
          "where": "README, image resources/bench_53.png, Agents' Last Exam panel, GLM-5.3-Flash bar"
        },
        {
          "url": "https://docs.z.ai/guides/llm/glm-5.3-flash",
          "name": "GLM-5.3-Flash card",
          "title": "GLM-5.3-Flash/FlashX - Overview",
          "lab": "zai",
          "date": "2026-08-26",
          "printed": "26.3",
          "where": "Benchmark image, panel 'Agents' Last Exam', GLM-5.3-Flash bar"
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/58a995e241d3e40c.png"
    },
    {
      "id": "qwen3-8-flash-next:agents-last-exam:pass:1",
      "model": "qwen3-8-flash-next",
      "model_name": "Qwen3.8-Flash-Next",
      "model_date": "2026-08-26",
      "event": "https://frontierlog.ai/event/2026-08-26-qwen3-8-flash-next",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 24.3,
      "unit": "%",
      "printed": "24.3",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.8-Flash-Next",
      "source": "https://qwen.ai/blog?id=qwen3.8-flash-next",
      "source_title": "Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency",
      "source_date": "2026-08-26",
      "where": "Language table (first table), Agent block, row 'Frontier agentic tasks / Agents' Last Exam', column 'Qwen3.8-Flash-Next', the value labelled 'Pass@1'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "pass@1; harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
          "name": "Qwen3.8-Flash-Next card",
          "title": "Qwen/Qwen3.8-Flash-Next model card",
          "lab": "alibaba",
          "date": "2026-08-26",
          "printed": "24.3",
          "where": "model card benchmark table, row \"Frontier agentic tasks / Agents' Last Exam\", column \"Qwen3.8-Flash-Next\", sub-value labelled \"Pass@1\""
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/1ce175beb0b20cf1.png"
    },
    {
      "id": "qwen3-8-flash-next:agents-last-exam:score:1",
      "model": "qwen3-8-flash-next",
      "model_name": "Qwen3.8-Flash-Next",
      "model_date": "2026-08-26",
      "event": "https://frontierlog.ai/event/2026-08-26-qwen3-8-flash-next",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "score",
      "setting_name": "Score",
      "score": 51.2,
      "unit": "%",
      "printed": "51.2",
      "reported_by": "alibaba",
      "run_by": null,
      "label": "Qwen3.8-Flash-Next",
      "source": "https://qwen.ai/blog?id=qwen3.8-flash-next",
      "source_title": "Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency",
      "source_date": "2026-08-26",
      "where": "Language table (first table), Agent block, row 'Frontier agentic tasks / Agents' Last Exam', column 'Qwen3.8-Flash-Next', the value labelled 'Score'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness and subset not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "url": "https://huggingface.co/Qwen/Qwen3.8-Flash-Next",
          "name": "Qwen3.8-Flash-Next card",
          "title": "Qwen/Qwen3.8-Flash-Next model card",
          "lab": "alibaba",
          "date": "2026-08-26",
          "printed": "51.2",
          "where": "model card benchmark table, row \"Frontier agentic tasks / Agents' Last Exam\", column \"Qwen3.8-Flash-Next\", sub-value labelled \"Score\""
        }
      ],
      "source_capture": "https://frontierlog.ai/captures/1ce175beb0b20cf1.png"
    },
    {
      "id": "gpt-6-astra:agents-last-exam: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "unstated",
      "setting_name": "Measure not stated",
      "score": 59.3,
      "unit": "%",
      "printed": "59.3%",
      "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": "'Computer Use' eval table, row 'Agents' Last Exam', column 'GPT-6 Astra'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "best of any effort; harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "deepseek-v4-1-flash:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 31.8,
      "unit": "%",
      "printed": "31.8",
      "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 'Agents' Last Exam', column 'DeepSeek V4.1-Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "CLI subset; pass@1; official ALE scaffold; max reasoning effort",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "corroborated_by": [
        {
          "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": "31.8",
          "where": "Tech report p.33, Table 3, row 'Agents’ Last Exam(Pass@1)', column 'DS-V4.1-Flash Max'; also prose 'Agents’ Last Exam (31.8%)'; sec 5.3.1 names the benchmark 'Agents’ Last Exam ... (ALE-CLI)', official scaffold"
        },
        {
          "url": "https://api-docs.deepseek.com/news/news261017/",
          "name": "DeepSeek-V4.1-Flash-1017",
          "title": "DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made Real",
          "lab": "deepseek",
          "date": "2026-10-17",
          "printed": "31.8",
          "where": "benchmark table image, row 'Agents' Last Exam', column 'DeepSeek V4.1-Flash'"
        },
        {
          "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": "31.8",
          "where": "Instruct Model, 'Comparison with frontier models (Max reasoning effort)' table, row 'Agent's Last Exam (Pass@1)', column 'DS-V4.1-Flash' (printed bold **31.8**); text says ALE uses its official scaffold, reasoning_effort=100"
        },
        {
          "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": "31.8",
          "where": "Introduction benchmark table, row 'Agents’ Last Exam', 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"
    },
    {
      "id": "gemini-4-argon:agents-last-exam:pass:1",
      "model": "gemini-4-argon",
      "model_name": "Gemini 4 Argon",
      "model_date": "2026-09-30",
      "event": "https://frontierlog.ai/event/2026-09-30-gemini-4-argon",
      "benchmark": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 39.5,
      "unit": "%",
      "printed": "39.5%",
      "reported_by": "google",
      "run_by": null,
      "label": "Gemini 4 Argon",
      "source": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/",
      "source_title": "Gemini 4 Argon: our next era of frontier intelligence",
      "source_date": "2026-09-30",
      "where": "benchmark table (image), row 'Computer use / Agent's Last Exam', 'Pass rate', column 'Gemini 4 Argon'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "ALE-Claw harness; 5-hour window; safety filters on",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": [],
      "source_image": "https://frontierlog.ai/sources/481de187c66d4b4d.gif",
      "source_image_original": "https://storage.googleapis.com/gweb-uniblog-publish-prod/original_images/gemini-4-argon_table_blog.gif"
    },
    {
      "id": "deepseek-v4-1-flash-1017:agents-last-exam:cli-pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "cli-pass",
      "setting_name": "CLI subset, pass rate",
      "score": 35.4,
      "unit": "%",
      "printed": "35.4",
      "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 'Agents' Last Exam', column 'DeepSeek V4.1-Flash-1017'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "harness not stated",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-29",
      "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": "35.4",
          "where": "Introduction benchmark table, row 'Agents’ Last Exam', column 'DS-V4.1-Flash-1017'"
        }
      ],
      "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:agents-last-exam:pass: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": "agents-last-exam",
      "benchmark_name": "Agents' Last Exam v1",
      "setting": "pass",
      "setting_name": "Pass rate",
      "score": 29.6,
      "unit": "%",
      "printed": "29.6",
      "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), Computer Use block, row 'Agents' Last Exam', column 'MiniMax M3.1 Flash'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "152 tasks, overall; Claude Code harness; max effort",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "double_entry",
      "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"
    }
  ]
}