UAE’s flagship LLM, Jais 2, is out: the next generation of the world’s leading Arabic open-weight LLM
Inception (a G42 company), Cerebras Systems, and the Institute of Foundation Models at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), yesterday announced Jais 2, the world’s leading open-weight Arabic Large Language Model (LLM). Built from the ground up with 70 billion parameters and trained on the richest Arabic-first dataset to date, the latest generation of Jais 2 will be available for download on Inception’s HuggingFace and the web application.
Powered by a redesigned architecture and cleaner training data, Jais delivers stronger reasoning, greater fluency across Modern Standard Arabic (MSA) and regional dialects, and strong performance in English. It is engineered to navigate real-world linguistic behavior, including code-switching and informal tone, while maintaining advanced technical and creative capabilities.
Jais 2 also integrates a comprehensive, safety-first framework supported by instruction tuning, evaluation, and continuous user feedback, ensuring the model remains reliable and adaptive over time. It has proven adept across a wide range of content, including Arabic poetry, culture, and social media tone of voice. The result is a model that captures Arabic as a language, as it is used every day across conversations, culture, and modern expression.
On the release, Ashish Koshy, CEO of Inception, commented: “The use of Arabic within AI has historically been constrained by limited datasets and fragmented representation. Jais 2 addresses these limitations by bringing unprecedented depth, nuance, and contextual intelligence to the language. From dialect to tone, the model understands Arabic as it is spoken, written, and lived. We are incredibly proud to help deliver a system that will accelerate innovation for millions across the region and beyond.”
Professor Preslav Nakov, Department Chair and Professor of Natural Language Processing, at MBZUAI, added: “Arabic has long been underserved in AI development due to limited high-quality data for training large language models. Today, with Jais, we mark a defining advancement for Arabic AI, as we share a model that is built not only with scale, but with cultural and linguistic fidelity at its core. By dramatically expanding the quality and diversity of Arabic data, we have created a foundation that reflects the richness of the Arabic language. This model stands as an example of how AI can evolve through cultural alignment, safety-centered design, and open innovation.”
