Inception, a G42 company, has advanced the field of artificial intelligence with the launch of JAIS 70B, a large language model (LLM) tailored for Arabic natural language processing (NLP).

This 70-billion parameter model is designed to enhance the integration of generative AI services across various sectors, including customer service, content creation, and data analysis.

JAIS 70B provides the open-source community with bilingual capabilities in Arabic and English. It was developed through continuous training and fine-tuned on 370 billion tokens, including the largest Arabic dataset ever used for an open-source foundational model, with 330 billion Arabic tokens.

Alongside JAIS 70B, Inception has launched 20 AI models ranging from 590 million to 70 billion parameters to address diverse needs, from small-scale applications to complex enterprise solutions. These models, fine-tuned for chat applications, have been trained on up to 1.6 trillion tokens in Arabic, English, and code. Dr. Andrew Jackson, CEO of Inception, highlighted the company’s dedication to preserving Arabic heritage and democratizing AI access through the JAIS model family.

“Releasing JAIS 70B and this new family of models reinforces our commitment to delivering the highest quality AI foundation model for Arabic-speaking nations,” he said.

Following the launch of JAIS-13B and JAIS-13B-chat in August 2023, Inception introduced JAIS-30B and JAIS-30B-chat, focusing on Arabic-centric models. JAIS 70B and JAIS 70B-chat have demonstrated superior performance in both English and Arabic benchmarks compared to earlier models.

Neha Sengupta, Principal Applied Scientist at Inception, noted that JAIS models up to 30 billion parameters were trained from scratch and outperformed adapted models. For the 70-billion parameter model, Inception built on Llama2 to enhance efficiency, reducing computational complexity and environmental impact while creating a more sustainable solution.

Now accepting applications for Fast Company Middle East’s Most Innovative Companies. Click here to apply.