G42 releases 70b parameter LLM for Arabic NLP

• JAIS 70B, built on Llama2 model, leverages an existing English knowledge base to create a more efficient and sustainable solution.
• Developed by Inception, the extensive suite of AI models sets a new benchmark in Arabic text processing and reasoning to empower the next generation of AI solutions. 

Inception, a G42 company specialising in advanced AI solutions, has released JAIS 70B, a 70 billion parametre large language model (LLM) designed specifically for Arabic natural language processing (NLP).

The release signifies a significant step forward in democratising access to AI for Arabic-speaking communities.

JAIS 70B boasts  Arabic-English bilingual capabilities for the open-source community. Trained on a massive dataset of 370 billion tokens, including 330 billion Arabic tokens, it surpasses previous models in its ability to handle complex tasks and datasets. The model’s development utilized continuous training techniques, fine-tuning a pre-trained model for optimal performance.

Inception has also launched a comprehensive suite of 20 JAIS foundation and fine-tuned models, ranging from 590 million to 70 billion parameters. These models are tailored for various applications, including chat, and offer diverse sizes to cater to both individual and enterprise needs.

The extensive release addresses a critical gap in the Arabic NLP landscape, providing developers with tools ranging from small, compute-efficient models for targeted applications to advanced models for enterprise-grade precision.

“JAIS 70B reinforces our commitment to delivering the highest quality AI foundation model for Arabic speaking nations,” Dr. Andrew Jackson, CEO of Inception, said.

“We are excited to extend our expertise in training and adapting models to other underserved languages.”

Building upon the Llama2 model, JAIS 70B leverages an existing English knowledge base to create a more efficient and sustainable solution. According to Inception, the model’s design significantly reduces the computational complexity and environmental impact associated with training from scratch.

With JAIS 70B, Inception continues to push the boundaries of Arabic-centric AI, providing a powerful tool for developers and researchers to advance NLP applications and empower Arabic-speaking communities.