{
  "api_version": "1",
  "updated": "2026-10-26",
  "latest_event": "2026-10-24",
  "rights": "© Frontier Log, all rights reserved.",
  "model": {
    "slug": "claude-opus-4-8",
    "name": "Claude Opus 4.8",
    "lab": "anthropic",
    "date": "2026-05-28",
    "source": "https://www.anthropic.com/news/claude-opus-4-8",
    "source_title": "Introducing Claude Opus 4.8",
    "flagship": true,
    "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
    "releases": [],
    "variant_of": null,
    "made_by": "Anthropic",
    "url": "https://frontierlog.ai/model/claude-opus-4-8",
    "results": [
      {
        "id": "claude-opus-4-8:hle:no-tools:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "no-tools",
        "setting_name": "No tools",
        "score": 49.8,
        "unit": "%",
        "printed": "49.8%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-4-8",
        "source_title": "Introducing Claude Opus 4.8",
        "source_date": "2026-05-28",
        "where": "benchmark table (image), row 'Multidisciplinary reasoning, Humanity's Last Exam', column 'Opus 4.8', 'no tools'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-16",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "49.8",
            "where": "Table 8.1.A (p.251), row Humanity's Last Exam / No tools, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
            "name": "Claude Opus 4.8 card",
            "title": "Claude Opus 4.8 System Card",
            "lab": "anthropic",
            "date": "2026-05-28",
            "printed": "49.8",
            "where": "Table 8.1.A (p.194), row Humanity's Last Exam / No tools, column Claude Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "49.8",
            "where": "Table 8.1.A (p.152), row Humanity's Last Exam / No tools, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-opus-5",
            "name": "Claude Opus 5",
            "title": "Introducing Claude Opus 5",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "49.8%",
            "where": "benchmark table image, row Humanity's Last Exam (no tools), column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-5",
            "name": "Claude Sonnet 5",
            "title": "Introducing Claude Sonnet 5",
            "lab": "anthropic",
            "date": "2026-06-30",
            "printed": "49.8%",
            "where": "benchmark table image, row Humanity's Last Exam (no tools), column Opus 4.8 (for reference)"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "49.8",
            "where": "Full Benchmark Table, row HLE-Full ; column Claude Opus 4.8 (max)"
          },
          {
            "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": "49.8",
            "where": "Chart 'LLM Performance Evaluation' (all models at max thinking effort), panel 'Humanity's Last Exam', bar 'Claude Opus 4.8' (lower label, not marked w/ Tools)"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png"
      },
      {
        "id": "claude-opus-4-8:hle:with-tools:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "hle",
        "benchmark_name": "Humanity's Last Exam",
        "setting": "with-tools",
        "setting_name": "With tools",
        "score": 57.9,
        "unit": "%",
        "printed": "57.9%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-4-8",
        "source_title": "Introducing Claude Opus 4.8",
        "source_date": "2026-05-28",
        "where": "benchmark table (image), row 'Multidisciplinary reasoning, Humanity's Last Exam', column 'Opus 4.8', 'with tools'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "57.9",
            "where": "Table 8.1.A (p.251), row Humanity's Last Exam / With tools, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
            "name": "Claude Opus 4.8 card",
            "title": "Claude Opus 4.8 System Card",
            "lab": "anthropic",
            "date": "2026-05-28",
            "printed": "57.9",
            "where": "Table 8.1.A (p.194), row Humanity's Last Exam / With tools, column Claude Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "57.9",
            "where": "Table 8.1.A (p.152), row Humanity's Last Exam / With tools, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-opus-5",
            "name": "Claude Opus 5",
            "title": "Introducing Claude Opus 5",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "57.9%",
            "where": "benchmark table image, row Humanity's Last Exam (with tools), column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-5",
            "name": "Claude Sonnet 5",
            "title": "Introducing Claude Sonnet 5",
            "lab": "anthropic",
            "date": "2026-06-30",
            "printed": "57.9%",
            "where": "benchmark table image, row Humanity's Last Exam (with tools), column Opus 4.8 (for reference)"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "57.9",
            "where": "Full Benchmark Table, row HLE-Full w/ tools ; column Claude Opus 4.8 (max)"
          },
          {
            "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": "57.9",
            "where": "Full Benchmark Table, row 'HLE w/ Tools (* = full set for Opus 4.8, GPT-5.5, Gemini 3.1 Pro)', column 'Claude Opus 4.8'"
          },
          {
            "url": "https://z.ai/blog/glm-5.3",
            "name": "GLM-5.3",
            "title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities",
            "lab": "zai",
            "date": "2026-08-14",
            "printed": "57.9",
            "where": "'Performance across comparison models' table, row 'HLE w/ Tools', column 'Opus 4.8'"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png"
      },
      {
        "id": "claude-opus-4-8:swe-bench-verified:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "swe-bench-verified",
        "benchmark_name": "SWE-bench Verified",
        "setting": "default",
        "setting_name": "",
        "score": 88.6,
        "unit": "%",
        "printed": "88.6",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Claude Opus 4.8",
        "source": "https://www.anthropic.com/claude-opus-4-8-system-card",
        "source_title": "Claude Opus 4.8 System Card",
        "source_date": "2026-05-28",
        "where": "Section 8.1 evaluation summary table (page 194), row 'SWE-bench Verified', column 'Claude Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "88.6",
            "where": "Table 8.1.A (p.251), row SWE-bench Verified, column Opus 4.8"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8ad3877c07561076.png"
      },
      {
        "id": "claude-opus-4-8:gpqa-diamond:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "gpqa-diamond",
        "benchmark_name": "GPQA Diamond",
        "setting": "default",
        "setting_name": "",
        "score": 93.6,
        "unit": "%",
        "printed": "93.6",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Claude Opus 4.8",
        "source": "https://www.anthropic.com/claude-opus-4-8-system-card",
        "source_title": "Claude Opus 4.8 System Card",
        "source_date": "2026-05-28",
        "where": "Section 8.1 evaluation summary table (page 194), row 'GPQA Diamond', column 'Claude Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "averaged over 25 trials",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://z.ai/blog/glm-5.2",
            "name": "GLM-5.2",
            "title": "GLM-5.2: Built for Long-Horizon Tasks",
            "lab": "zai",
            "date": "2026-06-16",
            "printed": "93.6",
            "where": "Full Benchmark Table, row 'GPQA-Diamond', column 'Claude Opus 4.8'"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8ad3877c07561076.png"
      },
      {
        "id": "claude-opus-4-8:browsecomp:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "browsecomp",
        "benchmark_name": "BrowseComp",
        "setting": "default",
        "setting_name": "",
        "score": 84.3,
        "unit": "%",
        "printed": "84.3",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Claude Opus 4.8",
        "source": "https://www.anthropic.com/claude-opus-4-8-system-card",
        "source_title": "Claude Opus 4.8 System Card",
        "source_date": "2026-05-28",
        "where": "Section 8.1 evaluation summary table (page 194), row 'BrowseComp', column 'Claude Opus 4.8', the '(single-agent)' value",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "single agent; adaptive thinking, max effort; 10M-token limit, compaction at 200K tokens",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "84.3",
            "where": "Table 8.1.A (p.251), row BrowseComp, column Opus 4.8 (single-agent)"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "84.3",
            "where": "Table 8.1.A (p.152), row BrowseComp, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-opus-5",
            "name": "Claude Opus 5",
            "title": "Introducing Claude Opus 5",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "84.3%",
            "where": "benchmark table image, row BrowseComp, column Opus 4.8"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "84.3",
            "where": "Full Benchmark Table, row BrowseComp ; column Claude Opus 4.8 (max)"
          },
          {
            "url": "https://openai.com/index/gpt-5-6",
            "name": "GPT-5.6",
            "title": "GPT-5.6: Frontier intelligence that scales with your ambition",
            "lab": "openai",
            "date": "2026-07-09",
            "printed": "84.3%",
            "where": "table 3 \"BrowseComp\" row, \"Claude Opus 4.8\" column"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/8ad3877c07561076.png"
      },
      {
        "id": "claude-opus-4-8:swe-bench-pro:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "swe-bench-pro",
        "benchmark_name": "SWE-bench Pro",
        "setting": "default",
        "setting_name": "",
        "score": 69.2,
        "unit": "%",
        "printed": "69.2%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-4-8",
        "source_title": "Introducing Claude Opus 4.8",
        "source_date": "2026-05-28",
        "where": "benchmark table (image), row 'Agentic coding, SWE-Bench Pro', column 'Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "The source doesn't say whether this is the public or the commercial set",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "69.2",
            "where": "Table 8.1.A (p.251), row SWE-bench Pro, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
            "name": "Claude Opus 4.8 card",
            "title": "Claude Opus 4.8 System Card",
            "lab": "anthropic",
            "date": "2026-05-28",
            "printed": "69.2",
            "where": "Table 8.1.A (p.194), row SWE-bench Pro, column Claude Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "69.2",
            "where": "Table 8.1.A (p.152), row SWE-bench Pro, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-5",
            "name": "Claude Sonnet 5",
            "title": "Introducing Claude Sonnet 5",
            "lab": "anthropic",
            "date": "2026-06-30",
            "printed": "69.2%",
            "where": "benchmark table image, row SWE-bench Pro, column Opus 4.8 (for reference)"
          },
          {
            "url": "https://openai.com/index/gpt-5-6",
            "name": "GPT-5.6",
            "title": "GPT-5.6: Frontier intelligence that scales with your ambition",
            "lab": "openai",
            "date": "2026-07-09",
            "printed": "69.2%",
            "where": "table 1 \"SWE-Bench Pro\" row, \"Claude Opus 4.8\" column"
          },
          {
            "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": "69.2",
            "where": "Performance chart image (performance.png), panel SWE-Pro, value label on the Opus4.8 bar"
          },
          {
            "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": "69.2",
            "where": "Full Benchmark Table, row 'SWE-bench Pro', column 'Claude Opus 4.8'"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png"
      },
      {
        "id": "claude-opus-4-8:deepswe-1-1:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "deepswe-1-1",
        "benchmark_name": "DeepSWE v1.1",
        "setting": "default",
        "setting_name": "",
        "score": 59.0,
        "unit": "%",
        "printed": "59.0%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-5",
        "source_title": "Introducing Claude Opus 5",
        "source_date": "2026-07-24",
        "where": "benchmark table (image), row 'Agentic coding, DeepSWE v1.1', column 'Opus 4.8'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-17",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "59.0",
            "where": "Table 8.1.A (p.152), row DeepSWE v1.1, column Opus 4.8"
          },
          {
            "url": "https://www.kimi.ai/blog/kimi-k3",
            "name": "Kimi K3",
            "title": "Kimi K3: Open Frontier Intelligence",
            "lab": "moonshot",
            "date": "2026-07-16",
            "printed": "59.0",
            "where": "Full Benchmark Table, row DeepSWE (footnote: DeepSWE v1.1 tasks) ; column Claude Opus 4.8 (max)"
          },
          {
            "url": "https://openai.com/index/gpt-5-6",
            "name": "GPT-5.6",
            "title": "GPT-5.6: Frontier intelligence that scales with your ambition",
            "lab": "openai",
            "date": "2026-07-09",
            "printed": "59%",
            "where": "table 1 \"DeepSWE v1.1\" row, \"Claude Opus 4.8\" column"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/924165396ca89130.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a8fb4f77a9fe240e6f27f3bdc47a137f3c74a29d-2600x2578.png"
      },
      {
        "id": "claude-opus-4-8:terminal-bench-2-1:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "terminal-bench-2-1",
        "benchmark_name": "Terminal-Bench 2.1",
        "setting": "default",
        "setting_name": "",
        "score": 74.6,
        "unit": "%",
        "printed": "74.6%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-4-8",
        "source_title": "Introducing Claude Opus 4.8",
        "source_date": "2026-05-28",
        "where": "benchmark table image, row 'Agentic terminal coding / Terminal-Bench 2.1', column 'Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "Terminus 2 harness",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
            "name": "Claude Opus 4.8 card",
            "title": "Claude Opus 4.8 System Card",
            "lab": "anthropic",
            "date": "2026-05-28",
            "printed": "74.6",
            "where": "Table 8.1.A (p.194), row Terminal-Bench 2.1, column Claude Opus 4.8"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png"
      },
      {
        "id": "claude-opus-4-8:osworld-verified:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "osworld-verified",
        "benchmark_name": "OSWorld-Verified",
        "setting": "default",
        "setting_name": "",
        "score": 83.4,
        "unit": "%",
        "printed": "83.4%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-4-8",
        "source_title": "Introducing Claude Opus 4.8",
        "source_date": "2026-05-28",
        "where": "benchmark table image, row 'Agentic computer use / OSWorld-Verified', column 'Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "run method changed; may not be comparable with earlier Anthropic OSWorld-Verified scores",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-18",
        "check": "human",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "83.4",
            "where": "Table 8.1.A (p.251), row OSWorld-Verified, column Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-opus-4-8-system-card",
            "name": "Claude Opus 4.8 card",
            "title": "Claude Opus 4.8 System Card",
            "lab": "anthropic",
            "date": "2026-05-28",
            "printed": "83.4",
            "where": "Table 8.1.A (p.194), row OSWorld-Verified, column Claude Opus 4.8"
          },
          {
            "url": "https://www.anthropic.com/claude-sonnet-5-system-card",
            "name": "Claude Sonnet 5 card",
            "title": "Claude Sonnet 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-30",
            "printed": "83.4%",
            "where": "Figure 8.10.2.A External OSWorld-Verified scores (p.126), bar Claude Opus 4.8 value label"
          },
          {
            "url": "https://www.anthropic.com/news/claude-sonnet-5",
            "name": "Claude Sonnet 5",
            "title": "Introducing Claude Sonnet 5",
            "lab": "anthropic",
            "date": "2026-06-30",
            "printed": "83.4%",
            "where": "benchmark table image, row OSWorld-Verified, column Opus 4.8 (for reference)"
          },
          {
            "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": "83.4",
            "where": "Performance chart image (performance.png), panel OSWorld-Verified, value label on the Opus4.8 bar"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/7ad9b784f2579c8a.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a9007019094f217e98cb8261a2765d7646c01708-2600x1392.png"
      },
      {
        "id": "claude-opus-4-8:arc-agi-2:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "arc-agi-2",
        "benchmark_name": "ARC-AGI-2",
        "setting": "default",
        "setting_name": "",
        "score": 72.1,
        "unit": "%",
        "printed": "72.1",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/claude-opus-5-system-card",
        "source_title": "Claude Opus 5 System Card",
        "source_date": "2026-07-24",
        "where": "Section 8.1, Table 8.1.A (page 152), row 'ARC-AGI-2', column 'Opus 4.8'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-19",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_capture": "https://frontierlog.ai/captures/f64941f07dfe6a79.png"
      },
      {
        "id": "claude-opus-4-8:automationbench:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "automationbench",
        "benchmark_name": "AutomationBench",
        "setting": "default",
        "setting_name": "",
        "score": 15.5,
        "unit": "%",
        "printed": "15.5",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Claude Opus 4.8",
        "source": "https://www.anthropic.com/claude-opus-4-8-system-card",
        "source_title": "Claude Opus 4.8 System Card",
        "source_date": "2026-05-28",
        "where": "Section 8.1, Table 8.1.A (evaluation summary), row 'Automation Bench', column 'Claude Opus 4.8'",
        "own": true,
        "primary": true,
        "parallel": false,
        "notes": "Zapier leaderboard, private held-out set; max effort",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-fable-5-mythos-5-system-card",
            "name": "Claude Fable 5 and Mythos 5 card",
            "title": "Claude Fable 5 and Claude Mythos 5 System Card",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "15.5",
            "where": "Table 8.1.A, row AutomationBench, column Claude Opus 4.8 (also 8.17.9 text: 'Claude Opus 4.8 (max effort) at 15.5%')"
          },
          {
            "url": "https://www.anthropic.com/news/claude-fable-5-mythos-5",
            "name": "Claude Fable 5 and Mythos 5",
            "title": "Claude Fable 5 and Claude Mythos 5",
            "lab": "anthropic",
            "date": "2026-06-09",
            "printed": "15.5%",
            "where": "Benchmark table image, row Tool use AutomationBench, column Claude Opus 4.8"
          }
        ],
        "source_capture": "https://frontierlog.ai/captures/37133ade6e699bde.png"
      },
      {
        "id": "claude-opus-4-8:automationbench:default:2",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "automationbench",
        "benchmark_name": "AutomationBench",
        "setting": "default",
        "setting_name": "",
        "score": 17.0,
        "unit": "%",
        "printed": "17.0%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-5",
        "source_title": "Introducing Claude Opus 5",
        "source_date": "2026-07-24",
        "where": "benchmark table (image), row 'Business workflows, AutomationBench', column 'Opus 4.8'",
        "own": false,
        "primary": false,
        "parallel": false,
        "notes": "Zapier leaderboard, private held-out set; max effort",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "17.0",
            "where": "Table 8.1.A, row AutomationBench, column Opus 4.8; 8.13.7 text: 'Claude Opus 4.8 (max effort) at 17.0%'"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/924165396ca89130.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a8fb4f77a9fe240e6f27f3bdc47a137f3c74a29d-2600x2578.png"
      },
      {
        "id": "claude-opus-4-8:arc-agi-3:default:1",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "arc-agi-3",
        "benchmark_name": "ARC-AGI-3",
        "setting": "default",
        "setting_name": "",
        "score": 1.5,
        "unit": "%",
        "printed": "1.5%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-opus-5",
        "source_title": "Introducing Claude Opus 5",
        "source_date": "2026-07-24",
        "where": "benchmark table (image), row 'Novel problem-solving, ARC-AGI-3', column 'Opus 4.8'",
        "own": false,
        "primary": true,
        "parallel": false,
        "notes": "verified by the ARC Prize Foundation; high effort; semi-private set",
        "caveat": "",
        "human_verified": false,
        "human_verified_on": null,
        "check": "corroborated",
        "corroborated": true,
        "corroborated_on": "2026-10-19",
        "corroborated_by": [
          {
            "url": "https://www.anthropic.com/claude-opus-5-system-card",
            "name": "Claude Opus 5 card",
            "title": "Claude Opus 5 System Card",
            "lab": "anthropic",
            "date": "2026-07-24",
            "printed": "1.5",
            "where": "Table 8.1.A, row ARC-AGI-3, column Opus 4.8 (8.14.2 text: 'Claude Opus 4.8 reached 1.52% at high effort')"
          }
        ],
        "source_image": "https://frontierlog.ai/sources/924165396ca89130.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/a8fb4f77a9fe240e6f27f3bdc47a137f3c74a29d-2600x2578.png"
      },
      {
        "id": "claude-opus-4-8:terminal-bench-2-1:default:2",
        "model": "claude-opus-4-8",
        "model_name": "Claude Opus 4.8",
        "model_date": "2026-05-28",
        "event": "https://frontierlog.ai/event/2026-05-28-claude-opus-4-8",
        "benchmark": "terminal-bench-2-1",
        "benchmark_name": "Terminal-Bench 2.1",
        "setting": "default",
        "setting_name": "",
        "score": 82.7,
        "unit": "%",
        "printed": "82.7%",
        "reported_by": "anthropic",
        "run_by": null,
        "label": "Opus 4.8",
        "source": "https://www.anthropic.com/news/claude-sonnet-5",
        "source_title": "Introducing Claude Sonnet 5",
        "source_date": "2026-06-30",
        "where": "benchmark table (image), row 'Agentic coding, Terminal-Bench 2.1', column 'Opus 4.8 (For reference)'",
        "own": false,
        "primary": false,
        "parallel": false,
        "notes": "",
        "caveat": "",
        "human_verified": true,
        "human_verified_on": "2026-10-20",
        "check": "human",
        "corroborated": false,
        "corroborated_on": null,
        "corroborated_by": [],
        "source_image": "https://frontierlog.ai/sources/849d6a908f0d46ac.png",
        "source_image_original": "https://www-cdn.anthropic.com/images/4zrzovbb/website/9941d610909f28a504e16dd5af823df172ec6035-2600x1234.png"
      }
    ]
  }
}