Inception, a subsidiary of G42 and a leader in developing advanced AI models and applications has announced the launch of its latest large language model, “Jais 70B,” along with 20 other AI models aimed at supporting Arabic natural language processing (NLP).
The company revealed that the “Jais 70B” model, with its 70 billion parameters, is designed specifically for developers working on Arabic NLP solutions.
This model promises to accelerate the integration of generative AI services across various sectors, thereby enhancing capabilities in areas such as customer service, content creation, and data analysis.
Moreover, the company unveiled a comprehensive suite of 20 finely-tuned foundational “Jais” models, spanning eight sizes.
These models range from 590 million to 70 billion parameters and are meticulously trained for conversational applications.
Furthermore, they have been trained on up to 1.6 trillion tokens in Arabic, English, and coding data.
The “Jais” models are designed to support a wide range of applications and aim to accelerate innovation, development, and research opportunities for future applications tailored to the Arabic-speaking and bilingual communities.
Dr. Andrew Jackson, CEO of Inception, stated, “Today, AI has become a significant value-adding force, and large language models have been at the forefront of the increased adoption of AI.”
Moreover, he emphasised that “Jais” was created to preserve the Arabic heritage, culture, and language and to make AI accessible to everyone.
He added, “The release of Jais 70B and this new family of models reinforces our commitment to providing a high-quality foundational AI model for Arab countries. The training and adaptation techniques we successfully provide for Arabic language models are scalable to include other underserved languages.”
Meanwhile, Neha Sengupta, a principal applied scientist at Inception, noted, “For models up to 30 billion parameters, we have successfully trained Jais from scratch, consistently outperforming modified models in the community. However, for models with 70 billion parameters and above, the computational complexities and environmental impact of training from scratch were significant.”
These advancements underscore G42’s commitment to driving forward the capabilities of AI, particularly in serving the needs of the Arabic-speaking world and beyond.
