OpenGCUA macOS four-task completion: leaderboard
OpenGCUA real-device macOS harness, four tasks, selected run1 per task; task-level capability, tiny preliminary cohort, not a general ranking claim. Workbook, PDF, shopping research and hotel comparison. DeepSeek omitted by source because hotel trajectory is missing; no invented aggregate. Metric: Complete tasks (out of 4). Higher is better. Source: https://arxiv.org/html/2610.22200v1
Metric: Complete tasks (out of 4). Source: arxiv.org. Saturation forecast: Around January 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 4 | #5 |
| 2 | Claude Opus 5 | 3 | #12 |
| 3 | Qwen 3.8 Max | 3 | #31 |
| 4 | Gemini 3.8 Flash | 2 | #21 |
| 5 | Qwen 3.8 Flash | 2 | #47 |
No result here: #3 Claude Opus 5.5, #7 Gemini 4 Argon, #8 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opengcua-macos-four-task-completion · How It Works · Data refreshed daily, snapshot 2026-10-17.