Core42 sets new benchmark for Arabic LLM with Jais 30B

Abu Dhabi artificial intelligence (AI) firm G42 Group’s subsidiary, Core42, has launched Jais 30B, the latest and most proficient advanced Arabic language software that can power generative AI applications.

The new iteration of Jais, which features 30 billion parameters, follows the release of the 13 billion parameter model in August. It is part of Core42’s broader strategy to provide a rich linguistic and culture-focused generative AI experience for more than 400 million Arabic speakers worldwide.

The new Jais 30B model was trained on a larger dataset than its predecessor, made of 126 billion Arabic tokens, 251 billion English tokens, and 50 billion code tokens and shows an increased performance across all key indicators.

The homegrown large language model (LLM) offers 160 per cent longer and more detailed answers in Arabic and a 233 per cent increase in English, reflecting significant improvements in language generation.

“Jais 30B performance is now on par with monolingual English models and outperforms most open-source models in Foundation Model evaluations,” Core42 said in a statement.

Expanding the horizon

The cloud and generative AI firm said Jais 30B’s enhancements have been tested and validated using heuristic, cross-model comparison, and human evaluations.

The results revealed that the responses of the model’s “fine-tuned” iterations outperform those of Jais 13B by 96 per cent of the time in Arabic and 97 per cent in English.

“The launch of Jais 30B marks another significant milestone for Core42 and represents a giant leap forward for the Arabic-speaking world in harnessing the potential of generative AI,” said Dr. Andrew Jackson, the EVP & chief AI officer of Core42.

“This release underscores the powerful synergy between Core42’s technological leadership, our extensive partner ecosystem, and our shared dedication to pushing the boundaries of what’s possible in the field of AI.”

Get ready to meet Jais 30B – the latest and most sophisticated addition to our open-source Arabic Large Language Model (#LLM) family!

With 30 billion parameters, it elevates Arabic verbosity by 160% and English by 233%, setting new language generation benchmarks. pic.twitter.com/6gs3AyePMF

— Core42 (@core42_ai) November 9, 2023

Named after the highest peak in the UAE, Jais is a collaboration between G42’s Inception, Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and Cerebras Systems.

Jais is expected to facilitate and expedite innovation while cementing Abu Dhabi’s position as a hub for AI, innovation, culture preservation and international collaboration.

Jais was created with the help of Condor Galaxy 1, the supercomputer produced by Silicon Valley-based Cerebras Systems, which designs dinner plate-sized chips. CG1 has the capacity of 4 exaFLOPs and 54 million cores, with 64 Cerebras CS-2 systems linked together into a single, easy-to-use AI supercomputer.

CG1 is part of a network of nine interconnected supercomputers that were unveiled by G42 and Cerebras Systems in July. The supercomputers offer a new approach to AI computing that promises to significantly reduce model training time.

Building Core42

Meanwhile, G42 merged three key subsidiaries – G42 Cloud, Inception, and Injazat – to create Core42 earlier in Core42 to focus on delivering AI solutions and services on a national scale.

Core42, which is being led by Kiril Evtimov, is combining top-tier technology, innovation, and services in a first-of-its-kind offering that includes scalable cloud solutions, AI cloud and HPC infrastructure catering to the public sector and regulated industries.

The company offers cutting-edge applied AI research and implementation, focusing on enabling generative AI for enterprise use cases across multiple sectors and government organisations.

It also provides cybersecurity services to protect organisations against data breaches and cyberattacks.