{
  "api_version": "1",
  "updated": "2026-10-20",
  "latest_event": "2026-10-18",
  "rights": "© Frontier Log, all rights reserved.",
  "benchmark": {
    "slug": "osworld-2-1",
    "name": "OSWorld 2.1",
    "family": "osworld",
    "unit": "%",
    "higher_is_better": true,
    "status": "active",
    "active": true,
    "superseded_by": null,
    "benchmark_url": "https://github.com/xlang-ai/OSWorld-V2/releases/tag/osworld-v2.1",
    "introduced": "2026-09-16",
    "creator": "XLANG Lab",
    "introduced_source": "https://github.com/xlang-ai/OSWorld-V2/releases/tag/osworld-v2.1",
    "settings": [
      {
        "setting": "partial",
        "name": "Full set, partial score",
        "url": "https://frontierlog.ai/benchmark/osworld-2-1",
        "results": 5
      },
      {
        "setting": "strict",
        "name": "Full set, strict pass rate",
        "url": "https://frontierlog.ai/benchmark/osworld-2-1/strict",
        "results": 3
      },
      {
        "setting": "offline-partial",
        "name": "Offline set, partial score",
        "url": "https://frontierlog.ai/benchmark/osworld-2-1/offline-partial",
        "results": 4
      },
      {
        "setting": "offline-strict",
        "name": "Offline set, strict pass rate",
        "url": "https://frontierlog.ai/benchmark/osworld-2-1/offline-strict",
        "results": 3
      }
    ]
  },
  "total": 15,
  "results": [
    {
      "id": "claude-haiku-4-5:osworld-2-1:offline-partial:1",
      "model": "claude-haiku-4-5",
      "model_name": "Claude Haiku 4.5",
      "model_date": "2025-10-15",
      "event": "https://frontierlog.ai/event/2025-10-15-claude-haiku-4-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-partial",
      "setting_name": "Offline set, partial score",
      "score": 15.7,
      "unit": "%",
      "printed": "15.7%",
      "reported_by": "anthropic",
      "label": "Haiku 4.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5",
      "source_title": "Introducing Claude Haiku 5.5",
      "source_date": "2026-10-07",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Haiku 4.5', the value tagged 'Offline subset'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; fixed thinking budget, no server-side context compaction; mean of 5 attempts",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-haiku-5-5-system-card",
          "name": "Claude Haiku 5.5 card",
          "title": "Claude Haiku 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-10-07",
          "printed": "15.7",
          "where": "8.1 Evaluation summary, row 'OSWorld 2.1 (offline subset)', column Haiku 4.5; also Figure 8.9.3.A bar label 15.7 (Claude Haiku 4.5, Partial)"
        }
      ]
    },
    {
      "id": "claude-sonnet-5:osworld-2-1:partial:1",
      "model": "claude-sonnet-5",
      "model_name": "Claude Sonnet 5",
      "model_date": "2026-06-30",
      "event": "https://frontierlog.ai/event/2026-06-30-claude-sonnet-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "partial",
      "setting_name": "Full set, partial score",
      "score": 57.0,
      "unit": "%",
      "printed": "57.0%",
      "reported_by": "anthropic",
      "label": "Sonnet 5",
      "source": "https://www.anthropic.com/claude-sonnet-5-5",
      "source_title": "Introducing Claude Sonnet 5.5",
      "source_date": "2026-09-28",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Sonnet 5', the value tagged 'partial'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
          "name": "Claude Sonnet 5.5 card",
          "title": "Claude Sonnet 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-28",
          "printed": "57.0%",
          "where": "section 8.13.3 OSWorld 2.1 (p. 130), sentence: 'Claude Sonnet 5 scored 57.0% partial and 25.6% strict pass'; also 8.1 Evaluation summary, row 'OSWorld 2.1', column Claude Sonnet 5: '57.0'"
        }
      ]
    },
    {
      "id": "claude-sonnet-5:osworld-2-1:strict:1",
      "model": "claude-sonnet-5",
      "model_name": "Claude Sonnet 5",
      "model_date": "2026-06-30",
      "event": "https://frontierlog.ai/event/2026-06-30-claude-sonnet-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "strict",
      "setting_name": "Full set, strict pass rate",
      "score": 25.6,
      "unit": "%",
      "printed": "25.6%",
      "reported_by": "anthropic",
      "label": "Claude Sonnet 5",
      "source": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
      "source_title": "Claude Sonnet 5.5 System Card",
      "source_date": "2026-09-28",
      "where": "section 8.13.3 OSWorld 2.1, text (PDF page 130), the Claude Sonnet 5 strict pass rate",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-opus-5:osworld-2-1:partial:1",
      "model": "claude-opus-5",
      "model_name": "Claude Opus 5",
      "model_date": "2026-07-24",
      "event": "https://frontierlog.ai/event/2026-07-24-claude-opus-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "partial",
      "setting_name": "Full set, partial score",
      "score": 74.0,
      "unit": "%",
      "printed": "74.0%",
      "reported_by": "anthropic",
      "label": "Opus 5",
      "source": "https://www.anthropic.com/claude-opus-5-5",
      "source_title": "Introducing Claude Opus 5.5",
      "source_date": "2026-09-22",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Opus 5', the value tagged 'partial'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files; the system card labels it OSWorld 2.0",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-5-5-system-card",
          "name": "Claude Opus 5.5 card",
          "title": "Claude Opus 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-22",
          "printed": "74.0%",
          "where": "section 8.13.3 OSWorld 2.0 (p. 206), sentence: 'Claude Opus 5 scored 74.0% partial and 37.2% strict pass'; also 8.1 Evaluation summary, row 'OSWorld 2.0 (partial/strict)', column Claude Opus 5: '74.0/37.2'"
        }
      ]
    },
    {
      "id": "claude-fable-5-1:osworld-2-1:partial:1",
      "model": "claude-fable-5-1",
      "model_name": "Claude Fable 5.1",
      "model_date": "2026-09-01",
      "event": "https://frontierlog.ai/event/2026-09-01-claude-fable-5-1-mythos-5-1",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "partial",
      "setting_name": "Full set, partial score",
      "score": 80.7,
      "unit": "%",
      "printed": "80.7%",
      "reported_by": "anthropic",
      "label": "Fable 5.1",
      "source": "https://www.anthropic.com/claude-opus-5-5",
      "source_title": "Introducing Claude Opus 5.5",
      "source_date": "2026-09-22",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Fable 5.1', the value tagged 'partial'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files; the system card labels it OSWorld 2.0",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-5-5-system-card",
          "name": "Claude Opus 5.5 card",
          "title": "Claude Opus 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-22",
          "printed": "80.7%",
          "where": "section 8.13.3 OSWorld 2.0 (p. 206), sentence: 'Claude Fable 5.1 scored 80.7% partial and 42.8% strict pass'; also 8.1 Evaluation summary, row 'OSWorld 2.0 (partial/strict)', column Claude Fable 5.1: '80.7/42.8', and Figure 8.13.3.A bar label 80.7"
        }
      ]
    },
    {
      "id": "claude-opus-5-5:osworld-2-1:partial:1",
      "model": "claude-opus-5-5",
      "model_name": "Claude Opus 5.5",
      "model_date": "2026-09-22",
      "event": "https://frontierlog.ai/event/2026-09-22-claude-opus-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "partial",
      "setting_name": "Full set, partial score",
      "score": 81.8,
      "unit": "%",
      "printed": "81.8%",
      "reported_by": "anthropic",
      "label": "Opus 5.5",
      "source": "https://www.anthropic.com/claude-opus-5-5",
      "source_title": "Introducing Claude Opus 5.5",
      "source_date": "2026-09-22",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Opus 5.5', the value tagged 'partial'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files; the system card labels it OSWorld 2.0",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-opus-5-5-system-card",
          "name": "Claude Opus 5.5 card",
          "title": "Claude Opus 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-22",
          "printed": "81.8%",
          "where": "section 8.13.3 OSWorld 2.0 (p. 206), sentence: 'Claude Opus 5.5 achieved a partial score of 81.8% and a strict pass rate of 48.7%'; also 8.1 Evaluation summary, row 'OSWorld 2.0 (partial/strict)', column Claude Opus 5.5: '81.8/48.7'"
        },
        {
          "url": "https://www.anthropic.com/claude-sonnet-5-5",
          "name": "Claude Sonnet 5.5",
          "title": "Introducing Claude Sonnet 5.5",
          "lab": "anthropic",
          "date": "2026-09-28",
          "printed": "81.8%",
          "where": "Benchmark table, row 'Computer use / OSWorld 2.1', column Opus 5.5: '81.8%' with 'partial' under it"
        },
        {
          "url": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
          "name": "Claude Sonnet 5.5 card",
          "title": "Claude Sonnet 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-28",
          "printed": "81.8%",
          "where": "section 8.13.3 OSWorld 2.1 (p. 130), sentence: 'Claude Opus 5.5 scored 81.8% partial and 48.7% strict pass'; also 8.1 Evaluation summary, row 'OSWorld 2.1', column Claude Opus 5.5: '81.8'"
        }
      ]
    },
    {
      "id": "claude-opus-5-5:osworld-2-1:offline-partial:1",
      "model": "claude-opus-5-5",
      "model_name": "Claude Opus 5.5",
      "model_date": "2026-09-22",
      "event": "https://frontierlog.ai/event/2026-09-22-claude-opus-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-partial",
      "setting_name": "Offline set, partial score",
      "score": 87.2,
      "unit": "%",
      "printed": "87.2%",
      "reported_by": "anthropic",
      "label": "Claude Opus 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5-system-card",
      "source_title": "Claude Haiku 5.5 System Card",
      "source_date": "2026-10-07",
      "where": "section 8.9.3 OSWorld 2.1 offline subset, text (PDF page 128), the Claude Opus 5.5 partial score",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-opus-5-5:osworld-2-1:offline-strict:1",
      "model": "claude-opus-5-5",
      "model_name": "Claude Opus 5.5",
      "model_date": "2026-09-22",
      "event": "https://frontierlog.ai/event/2026-09-22-claude-opus-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-strict",
      "setting_name": "Offline set, strict pass rate",
      "score": 53.2,
      "unit": "%",
      "printed": "53.2%",
      "reported_by": "anthropic",
      "label": "Claude Opus 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5-system-card",
      "source_title": "Claude Haiku 5.5 System Card",
      "source_date": "2026-10-07",
      "where": "section 8.9.3 OSWorld 2.1 offline subset, text (PDF page 128), the Claude Opus 5.5 strict pass rate",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-opus-5-5:osworld-2-1:strict:1",
      "model": "claude-opus-5-5",
      "model_name": "Claude Opus 5.5",
      "model_date": "2026-09-22",
      "event": "https://frontierlog.ai/event/2026-09-22-claude-opus-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "strict",
      "setting_name": "Full set, strict pass rate",
      "score": 48.7,
      "unit": "%",
      "printed": "48.7%",
      "reported_by": "anthropic",
      "label": "Claude Opus 5.5",
      "source": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
      "source_title": "Claude Sonnet 5.5 System Card",
      "source_date": "2026-09-28",
      "where": "section 8.13.3 OSWorld 2.1, text (PDF page 130), the Claude Opus 5.5 strict pass rate",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-sonnet-5-5:osworld-2-1:partial:1",
      "model": "claude-sonnet-5-5",
      "model_name": "Claude Sonnet 5.5",
      "model_date": "2026-09-28",
      "event": "https://frontierlog.ai/event/2026-09-28-claude-sonnet-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "partial",
      "setting_name": "Full set, partial score",
      "score": 80.1,
      "unit": "%",
      "printed": "80.1%",
      "reported_by": "anthropic",
      "label": "Sonnet 5.5",
      "source": "https://www.anthropic.com/claude-sonnet-5-5",
      "source_title": "Introducing Claude Sonnet 5.5",
      "source_date": "2026-09-28",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Sonnet 5.5', the value tagged 'partial'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
          "name": "Claude Sonnet 5.5 card",
          "title": "Claude Sonnet 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-09-28",
          "printed": "80.1%",
          "where": "section 8.13.3 OSWorld 2.1 (p. 130), sentence: 'Claude Sonnet 5.5 achieved a partial score of 80.1% and a strict pass rate of 43.5%'; also 8.1 Evaluation summary, row 'OSWorld 2.1', column Claude Sonnet 5.5: '80.1'"
        }
      ]
    },
    {
      "id": "claude-sonnet-5-5:osworld-2-1:strict:1",
      "model": "claude-sonnet-5-5",
      "model_name": "Claude Sonnet 5.5",
      "model_date": "2026-09-28",
      "event": "https://frontierlog.ai/event/2026-09-28-claude-sonnet-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "strict",
      "setting_name": "Full set, strict pass rate",
      "score": 43.5,
      "unit": "%",
      "printed": "43.5%",
      "reported_by": "anthropic",
      "label": "Claude Sonnet 5.5",
      "source": "https://www.anthropic.com/claude-sonnet-5-5-system-card",
      "source_title": "Claude Sonnet 5.5 System Card",
      "source_date": "2026-09-28",
      "where": "section 8.13.3 OSWorld 2.1, text (PDF page 130), the Claude Sonnet 5.5 strict pass rate",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "max effort; mean of 5 runs; September 10, 2026 task files",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-sonnet-5-5:osworld-2-1:offline-partial:1",
      "model": "claude-sonnet-5-5",
      "model_name": "Claude Sonnet 5.5",
      "model_date": "2026-09-28",
      "event": "https://frontierlog.ai/event/2026-09-28-claude-sonnet-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-partial",
      "setting_name": "Offline set, partial score",
      "score": 83.9,
      "unit": "%",
      "printed": "83.9%",
      "reported_by": "anthropic",
      "label": "Sonnet 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5",
      "source_title": "Introducing Claude Haiku 5.5",
      "source_date": "2026-10-07",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Sonnet 5.5', the value tagged 'Offline subset'",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-haiku-5-5-system-card",
          "name": "Claude Haiku 5.5 card",
          "title": "Claude Haiku 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-10-07",
          "printed": "83.9%",
          "where": "section 8.9.3 OSWorld 2.1 offline subset (p. 128), sentence: 'Claude Sonnet 5.5 scored 83.9% partial and 48.8% strict pass'; also 8.1 Evaluation summary, row 'OSWorld 2.1 (offline subset)', column Sonnet 5.5: '83.9'"
        }
      ]
    },
    {
      "id": "claude-sonnet-5-5:osworld-2-1:offline-strict:1",
      "model": "claude-sonnet-5-5",
      "model_name": "Claude Sonnet 5.5",
      "model_date": "2026-09-28",
      "event": "https://frontierlog.ai/event/2026-09-28-claude-sonnet-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-strict",
      "setting_name": "Offline set, strict pass rate",
      "score": 48.8,
      "unit": "%",
      "printed": "48.8%",
      "reported_by": "anthropic",
      "label": "Claude Sonnet 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5-system-card",
      "source_title": "Claude Haiku 5.5 System Card",
      "source_date": "2026-10-07",
      "where": "section 8.9.3 OSWorld 2.1 offline subset, text (PDF page 128), the Claude Sonnet 5.5 strict pass rate",
      "own": false,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    },
    {
      "id": "claude-haiku-5-5:osworld-2-1:offline-partial:1",
      "model": "claude-haiku-5-5",
      "model_name": "Claude Haiku 5.5",
      "model_date": "2026-10-07",
      "event": "https://frontierlog.ai/event/2026-10-07-claude-haiku-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-partial",
      "setting_name": "Offline set, partial score",
      "score": 72.4,
      "unit": "%",
      "printed": "72.4%",
      "reported_by": "anthropic",
      "label": "Haiku 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5",
      "source_title": "Introducing Claude Haiku 5.5",
      "source_date": "2026-10-07",
      "where": "benchmark table (HTML), row 'Computer use / OSWorld 2.1', column 'Haiku 5.5', the value tagged 'Offline subset'",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": false,
      "human_verified_on": null,
      "check": "corroborated",
      "corroborated": true,
      "corroborated_on": "2026-10-20",
      "corroborated_by": [
        {
          "url": "https://www.anthropic.com/claude-haiku-5-5-system-card",
          "name": "Claude Haiku 5.5 card",
          "title": "Claude Haiku 5.5 System Card",
          "lab": "anthropic",
          "date": "2026-10-07",
          "printed": "72.4%",
          "where": "section 8.9.3 OSWorld 2.1 offline subset (p. 128), sentence: 'Claude Haiku 5.5 achieved a partial score of 72.4% and a strict pass rate of 37.1%'; also 8.1 Evaluation summary, row 'OSWorld 2.1 (offline subset)', column Haiku 5.5: '72.4'"
        }
      ]
    },
    {
      "id": "claude-haiku-5-5:osworld-2-1:offline-strict:1",
      "model": "claude-haiku-5-5",
      "model_name": "Claude Haiku 5.5",
      "model_date": "2026-10-07",
      "event": "https://frontierlog.ai/event/2026-10-07-claude-haiku-5-5",
      "benchmark": "osworld-2-1",
      "benchmark_name": "OSWorld 2.1",
      "setting": "offline-strict",
      "setting_name": "Offline set, strict pass rate",
      "score": 37.1,
      "unit": "%",
      "printed": "37.1%",
      "reported_by": "anthropic",
      "label": "Claude Haiku 5.5",
      "source": "https://www.anthropic.com/claude-haiku-5-5-system-card",
      "source_title": "Claude Haiku 5.5 System Card",
      "source_date": "2026-10-07",
      "where": "section 8.9.3 OSWorld 2.1 offline subset, text (PDF page 128), the Claude Haiku 5.5 strict pass rate",
      "own": true,
      "primary": true,
      "parallel": false,
      "notes": "82-task offline subset, no internet access; max effort; mean of 5 attempts",
      "caveat": "",
      "human_verified": true,
      "human_verified_on": "2026-10-20",
      "check": "human",
      "corroborated": false,
      "corroborated_on": null,
      "corroborated_by": []
    }
  ]
}