Advanced AI Technologies

Abu Dhabi’s new Arabic AI can understand dialects and retain English terms

Abu Dhabi-based MBZUAI and G42 launched Jais 3, an Arabic AI model family designed to understand dialects, retain English terms, use software tools, and complete multi-step tasks.

Mohamed bin Zayed University of Artificial Intelligence’s Institute of Foundation Models and G42’s Inception42 launched Jais 3 in Abu Dhabi. The model family combines Arabic speech recognition with instruction-following, software-tool use, and multi-step task execution. It is designed to understand Arabic dialects while retaining English terms commonly used in speech. Jais 3 reflects the region’s growing effort to develop Arabic-focused foundation models for practical applications. Its capabilities could support Arabic voice interfaces, enterprise automation, and agentic workflows. The launch also adds to a competitive regional ecosystem that includes Saudi Arabia’s ALLaM, Qatar’s Fanar, and the UAE’s Falcon Arabic. Developing models trained for local linguistic conditions can improve usability compared with systems optimized primarily for standard written Arabic or English. The model’s ability to use tools and perform sequences of tasks moves it beyond basic text generation toward operational AI agents. That makes the technology relevant to customer service, government services, and business software integration. The announcement underscores Abu Dhabi’s role in building locally developed AI capabilities and infrastructure.

Original source: Khaleej Times

Related RAQI solutions

In-depth analysis on RAQI Insight

Work with RAQI