Alibaba’s Tongyi Lab has open-sourced MAI-UI, a GUI agent framework, releasing the paper, code, and full-size fashions (2B/8B/32B/235B-A22B) defending edge to cloud deployment, enabling cross-app collaboration and privacy-protected interactions for AI terminals.
MAI-UI overcomes standard GUI agent limitations by actively querying clients for missing particulars and calling exterior APIs to streamline operations—equal to integrating Amap API for commute comparisons or GitHub API for commit extraction and emailing—with out information app switching. Its revolutionary end-cloud system dynamically assigns duties: privacy-sensitive operations hold native, superior ones go to the cloud, boosting the 2B edge model’s success cost by 33% and lowering cloud calls by over 40%, with larger than 40% of duties handled domestically for effectivity and security.
Effectivity highlights set enterprise knowledge: 76.7% success cost on AndroidWorld cellphone navigation (surpassing Gemini-2.5-Skilled), 91.3% on MMBench GUI L2 accuracy, and 73.5% on ScreenSpot-Skilled issue positioning, far outperforming mates. Even the smallest 2B edge model achieves 49.1% navigation success, a 75% enchancment over standard edge fashions.
MAI-UI is now completely open on GitHub and arXiv, empowering builders to deploy and velocity up human-like interactions on AI telephones and wise items.
Provide: QbitAI
Elevate your perspective with NextTech Info, the place innovation meets notion.
Uncover the latest breakthroughs, get distinctive updates, and be a part of with a worldwide group of future-focused thinkers.
Unlock tomorrow’s tendencies proper this second: be taught additional, subscribe to our e-newsletter, and alter into part of the NextTech group at NextTech-news.com
Keep forward of the curve with NextBusiness 24. Discover extra tales, subscribe to our publication, and be a part of our rising group at nextbusiness24.com
