Understand this story
On September 23, 2026, Qwen announced the performance of its Mobile-Use Agent. According to official information, the agent executes tasks with an API-first, GUI-fallback approach, scoring 82.1 on MobileWorld, 92.2 on MobileWorld-Real, and 97.2 on AndroidDaily, with an end-to-end success rate of 90%. In the same release, Qwen also introduced Qwen Intelligence and three mobile agents, and open-sourced the MobilePA-Bench, MobileWorld, MobileWorld-Real, and MobileWorld-Safety evaluation suites, covering planning, cross-app execution, real-device performance, and safety. There has been no subsequent independent development so far.
AI-generated from public reports
Latest development · Report
Qwen opens MobilePA-Bench, MobileWorld, and other mobile agent benchmark suitesDevelopments
1 confirmed developments
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Sep 23
Tongyi Qianwen / QwenFirst-party
Qwen opens MobilePA-Bench, MobileWorld, and other mobile agent benchmark suites
Qwen has open-sourced four benchmark suites—MobilePA-Bench, MobileWorld, MobileWorld-Real, and MobileWorld-Safety—covering planning, cross-app execution, real-device performance, and safety. These benchmarks correspond to the three types of agents released by Qwen Intelligence: Mobile Planner Agent ranks first on MobilePA-Bench, MobilePA-Bench Business, and Memory; Mobile-Use Agent scores 82.1 on MobileWorld, 92.2 on MobileWorld-Real, and 97.2 on AndroidDaily, with an end-to-end success rate of 90%; Mobile Creative Agent can generate images within 3 seconds, roughly twice as fast as comparable products.
alibaba_qwenFirst-party
English translation is being prepared.
alibaba_qwenFirst-party
English translation is being prepared.
alibaba_qwenFirst-party
English translation is being prepared.
alibaba_qwenFirst-partySelected
English translation is being prepared.