WindowsWorld (Set-of-Marks) - Intermediate Checkpoints: leaderboard

GUI agents on WindowsWorld (2026), 181 persona-driven professional tasks across 17 Windows desktop applications, mostly multi-application workflows plus infeasible tasks to reject, with the model observing Set-of-Marks annotated screenshots and acting through PyAutoGUI. Share of human-verified intermediate checkpoints reached on the feasible tasks, in percent; higher is better.

Metric: Intermediate checkpoint score (%): share of each task's human-verified intermediate checkpoints reached, averaged over the 169 feasible Level 1 to 3 tasks of WindowsWorld (persona-driven Windows desktop tasks across 17 applications, executed with a unified PyAutoGUI action space and level-specific step budgets of 15, 25, 40 and 20), the model observing Set-of-Marks annotated screenshots, states judged by a human-validated Qwen3-VL-Plus judge; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash (Preview)39.38#156
2Gemini 3 Pro (Preview)33.88#123
3Claude Sonnet 4.54.9#267
4GPT-5.24.57#190
5Qwen 3 VL Plus4.27#514

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=windowsworld-set-of-marks-intermediate-checkpoints · How It Works · Data refreshed daily, snapshot 2026-11-22.