Core42 Unveils Arabic Large Language Model, JAIS 30B Chat, On Microsoft Azure

In a significant leap forward for Arabic language technology, Core42, a subsidiary of the renowned G42, has unveiled its latest advance: the JAIS 30B Chat Arabic Large Language Model (LLM). This innovation is now accessible on Microsoft Azure, following an announcement by Microsoft's Chairman and CEO, Satya Nadella, at the commencement of Microsoft Build 2024. This initiative is set to revolutionize the deployment of advanced AI technologies across various industries by leveraging Microsoft Azure AI's Model-as-a-Service offering. The service not only provides a pre-trained model and infrastructure but also allows for customization to cater to the unique needs and specifications of Core42's models.

The introduction of JAIS via Microsoft Azure marks a significant milestone in making generative AI more accessible to the 422 million Arabic speakers worldwide. It opens up new avenues for connecting with end-users through localized solutions, exploring new market opportunities, and fostering multi-lingual driven innovation among developers. JAIS 30B Chat distinguishes itself by setting new industry benchmarks with its sophisticated generative AI capabilities, adept at handling cross-lingual conversations, retrieval augmented generation (RAG), summarization, content generation, and reasoning, specifically tailored for the Arabic language.

Talal Al Kaissi, the Chief Product & Global Partnerships Officer at Core42, emphasized the importance of this collaboration with Microsoft, highlighting it as a critical step towards ensuring equitable access to advanced AI technologies. Through Microsoft Azure, Core42 aims to empower people and organizations worldwide to harness the power of AI for a wide array of applications, thereby driving societal benefits and moving the needle in the global AI domain.

From Microsoft's perspective, Ali Dalloul, Vice President of Azure AI Customer Experience Engineering, expressed enthusiasm about adding JAIS 30B Chat to their Model-as-a-Service offering. This addition underlines Microsoft's dedication to offering a diverse and comprehensive model catalog, enabling developers to extend their capabilities and adhere to Microsoft's commitment to providing a variety of model options.

JAIS 30B Chat is poised to deliver transformative solutions across multiple sectors. By optimizing communication, service delivery, and workflow efficiency, it has the potential to significantly impact education through language support and content translation, enhance healthcare with improved patient management and translation services, and stimulate creativity and engagement in the media and entertainment industry through innovative content generation and summarization.

Trained on a whopping 1.63 trillion tokens—including a significant 475 billion Arabic tokens and supplemented by English tokens and code—JAIS 30B Chat's bilingual training dataset is designed to improve logical reasoning abilities in both languages. This makes JAIS 30B Chat a leading Arabic LLM for producing high-quality output and handling complex tasks, offering more efficient processing of large volumes of Arabic text compared to most English LLMs.

This collaboration between Core42 and Microsoft Azure is a testament to the continual efforts to address critical technology gaps for Arabic speakers and promote inclusivity. By tailoring technologies like JAIS 30B Chat to the unique characteristics of the Arabic language, Core42 and Microsoft are at the forefront of enhancing AI accessibility and utility for a significant portion of the global population.