{
  "data": {
    "evaluations": "https://www.anchorterminal.com/benchmark/lab/evaluation/",
    "harness": "https://github.com/Anchor-Terminal/anchorlab",
    "lab": "https://www.anchorterminal.com/api/v1/lab.json",
    "labEvals": "https://www.anchorterminal.com/api/v1/lab-evals.json",
    "listingsTested": 102,
    "modelsScored": 15,
    "page": "https://www.anchorterminal.com/benchmark/lab/"
  },
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/benchmark/lab/",
    "json": "https://www.anchorterminal.com/benchmark/lab/index.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/benchmark/lab/index.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/benchmark/lab/index.min.md"
  },
  "markdown": "The lab is one machine we run ourselves, beside the site's server. On it open-weight models try the tools we list and answer questions about this site, and what they find is published with its date.\n\n## What it does\n\n- Evaluations: models asked questions that have one right answer, run the same way for every model. Open-weight models are the first. https://www.anchorterminal.com/benchmark/lab/evaluation/index.md\n- Tool tests: a model is given only what a server says about its tools and writes programs to call each one, and the lab's gateway counts and times every call. What it found is on the listing, under Lab, and moves its Performance and Task success. https://www.anchorterminal.com/benchmark/index.md\n\n## The setup\n\n- The box: an AMD Ryzen AI Max+ 395 with 128 GB, running plain Debian, every model through llama.cpp from a build pinned by file hash.\n- The cell: a program a model writes runs with no network, and what it calls goes through the lab's gateway.\n- Nothing the lab reads is trusted, and the site checks the box's results one field at a time before it shows any.\n- The harness is open source (Go, standard library only, MIT): https://github.com/Anchor-Terminal/anchorlab\n\nJSON: https://www.anchorterminal.com/api/v1/lab.json and https://www.anchorterminal.com/api/v1/lab-evals.json\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": "Benchmark",
        "url": "https://www.anchorterminal.com/benchmark/"
      },
      {
        "name": "Lab",
        "url": ""
      }
    ],
    "description": "One machine where open-weight models try the tools Anchor Terminal lists and answer questions about the site. What the lab does, how it's set up, and where its results go.",
    "facts": [
      "15 models scored",
      "102 listings tried",
      "every run dated"
    ],
    "h1": "The lab: we run the tools and the models ourselves",
    "image": "https://www.anchorterminal.com/assets/og/benchmark-lab.png",
    "path": "/benchmark/lab/",
    "published": "",
    "section": "benchmark",
    "title": "The lab: we run the tools and the models ourselves | Anchor Terminal",
    "toc": null,
    "updated": "2026-11-22",
    "url": "https://www.anchorterminal.com/benchmark/lab/"
  },
  "tokens": {
    "markdown": 500,
    "slim": 380
  },
  "version": 1
}
