iOSWorld (Vision-Only) - Memory: leaderboard

iOSWorld is a mobile GUI-agent benchmark of 133 tasks across 26 purpose-built iOS apps sharing one user identity, spanning single-app, multi-app and long-horizon memory tasks. This board reports Task pass rate (%) on the 46 long-horizon memory tasks, vision-only; higher is better.

Metric: Task pass rate (%) on the 46 long-horizon memory tasks, vision-only; higher is better. Source: arxiv.org. Saturation forecast: Around November 2028. 6 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash21.7#164
2Claude Sonnet 4.613#161
3GPT-5.4 Mini10.9#389
4Claude Opus 4.68.7#102
5GPT-5.46.5#123
6Qwen 3.5 35B A3B4.3#482

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=iosworld-vision-only-memory · How It Works · Data refreshed daily, snapshot 2026-11-13.