{
  "data": {
    "area": "agent-tasks",
    "attribution": "METR, task-completion time horizons, CC BY 4.0",
    "cadence": "none (no longer updated)",
    "categories": [
      "inference",
      "open-weights"
    ],
    "competitor": "",
    "data": "https://metr.org/assets/benchmark_results_1_1.yaml",
    "direction": "higher",
    "floor": 0,
    "json": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/index.json",
    "licence": "CC-BY-4.0",
    "licenceUrl": "https://creativecommons.org/licenses/by/4.0/",
    "measures": "The length of software, machine learning and security task, timed on human experts, that an agent is predicted to finish half the time, from a logistic fit over more than a hundred tasks.",
    "method": "https://metr.org/time-horizons/",
    "models": 1,
    "name": "METR time horizons",
    "newest": "2026-09-08",
    "page": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/",
    "read": "2026-11-20",
    "results": 1,
    "runBy": "METR",
    "saturated": false,
    "slug": "metr-time-horizon",
    "status": "discontinued",
    "unit": "minutes at 50% success",
    "versions": [
      {
        "rows": [
          {
            "as": "gemini_3_1_pro",
            "ci": [
              233.5,
              694.8
            ],
            "listing": "vertex-ai",
            "model": "gemini-3-1-pro-preview",
            "name": "Gemini 3.1 Pro (preview)",
            "of": 1,
            "rank": 1,
            "runBy": "METR",
            "score": 384.1,
            "setting": {
              "scaffold": "triframe_inspect/triframe_agent"
            },
            "tier": "independent",
            "url": "https://metr.org/time-horizons/",
            "version": "METR-Horizon-v1.1"
          }
        ],
        "version": "METR-Horizon-v1.1"
      }
    ]
  },
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/",
    "json": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/index.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/index.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/index.min.md"
  },
  "markdown": "The length of software, machine learning and security task, timed on human experts, that an agent is predicted to finish half the time, from a logistic fit over more than a hundred tasks.\n\n- Score: minutes at 50% success, higher is better\n- Run by: METR\n- Status: Discontinued, newest result 8 September 2026\n- Updated: none (no longer updated)\n- Data: https://metr.org/assets/benchmark_results_1_1.yaml (CC-BY-4.0, https://creativecommons.org/licenses/by/4.0/), read 2026-11-20\n- Method: https://metr.org/time-horizons/\n\n## Results\n\n1 model we list, each at its best setting, best first.\n\n| # | Model | Score | Range | Who ran it | Setting | Date | Source |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| 1 | [Gemini 3.1 Pro (preview)](https://www.anchorterminal.com/tools/vertex-ai.md) | 384 min | 234 min to 695 min | Independent | scaffold triframe_inspect/triframe_agent |  | https://metr.org/time-horizons/ |\n\nA rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades.\n\nData: METR, task-completion time horizons, CC BY 4.0, CC-BY-4.0. All benchmarks: https://www.anchorterminal.com/benchmarks/index.md\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-22",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "1.0",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "Benchmarks",
        "url": "https://www.anchorterminal.com/benchmarks/"
      },
      {
        "name": "METR time horizons",
        "url": ""
      }
    ],
    "description": "METR time horizons results for 1 AI models we list, run by METR, with each result's setting, date and source.",
    "facts": [
      "1 models",
      "METR",
      "Discontinued"
    ],
    "h1": "METR time horizons",
    "image": "https://www.anchorterminal.com/assets/og/benchmarks-metr-time-horizon.png",
    "path": "/benchmarks/metr-time-horizon/",
    "published": "",
    "section": "benchmarks",
    "title": "METR time horizons results for the AI models we list | Anchor Terminal",
    "toc": null,
    "updated": "2026-11-21",
    "url": "https://www.anchorterminal.com/benchmarks/metr-time-horizon/"
  },
  "tokens": {
    "markdown": 500,
    "slim": 430
  },
  "version": 1
}
