{
  "generated": "2026-09-08",
  "benchmark": "screenspot-pro",
  "results": [
    {
      "system": "OS-Atlas-7B",
      "developer": "Shanghai AI Lab (OS-Copilot)",
      "value": 18.9,
      "date": "2025-04-04",
      "source": {
        "url": "https://arxiv.org/abs/2504.07981",
        "kind": "paper",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Best grounding model in the paper's Table 2; GPT-4o scored 0.8-0.9%."
      }
    },
    {
      "system": "GPT-5.2 Thinking",
      "developer": "OpenAI",
      "value": 86.3,
      "date": "2025-12-11",
      "source": {
        "url": "https://openai.com/index/introducing-gpt-5-2/",
        "kind": "developer-report",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "tools": true,
        "reasoning_effort": "xhigh",
        "notes": "Python tool enabled; OpenAI states scores are much lower without it. GPT-5.1 Thinking: 64.2% under the same setup."
      }
    },
    {
      "system": "Indeed-UI-32B-zoomin",
      "developer": "Intelligence Indeed",
      "value": 82.7,
      "date": "2026-08-13",
      "source": {
        "url": "https://gui-agent.github.io/grounding-leaderboard/",
        "kind": "official-leaderboard",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "notes": "Top entry of the official leaderboard on access date; zoom-in variant. Date is the leaderboard's 'last updated' stamp, not the submission date."
      }
    },
    {
      "system": "Claude Mythos 5.1",
      "developer": "Anthropic",
      "value": 87.3,
      "date": "2026-09-03",
      "source": {
        "url": "https://openai.com/index/gpt-6-astra/",
        "kind": "developer-report",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "tools": false,
        "notes": "Run by OpenAI in its GPT-6 Astra comparison table (column 'Claude Fable 5'); footnote says the score comes from Mythos, the reduced-safeguard variant. Not an Anthropic-reported number."
      }
    },
    {
      "system": "GPT-6 Astra",
      "developer": "OpenAI",
      "value": 92.7,
      "date": "2026-09-03",
      "source": {
        "url": "https://openai.com/index/gpt-6-astra/",
        "kind": "developer-report",
        "accessed": "2026-09-04"
      },
      "conditions": {
        "tools": false,
        "notes": "'ScreenSpot-Pro (no tools)' row; maximum effort. GPT-5.6 Sol: 76.9% in the same table."
      }
    }
  ]
}