G42, the Abu Dhabi-based global technology group, has announced a major upgrade to Llama-3-Nanda, or NANDA for short, its open-source Hindi-English large language model (LLM). Developed by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in collaboration with Inception, a G42 company, and Cerebras, AI inference makers, NANDA now features 87 billion parameters!
In an official statement, G42 revealed that NANDA 87B is built upon Llama-3.1 70B, and has been trained on a curated Hindi-English dataset with over 65 billion Hindi tokens. “A custom Hindi-centric tokenizer boosts efficiency, reducing both training and inference time” and that this breakthrough makes it one of the largest Hindi-centric models available in open weights.
The model is engineered for real-world use: fluent in formal Hindi (Devanagari script), casual speech, and Hinglish (a commonly used mish-mash of Hindi and English), and it aims to deliver strong performance across translation, summarization, instruction-following, and transliteration tasks. “Safety and cultural alignment are core to its design, enabling NANDA to generate context-aware, responsible responses,” says G42.
“India deserves world-class technology that speaks its language. NANDA 87B is a major step in that direction, after our first NANDA model announced last year,” said Manu Jain, CEO of G42 India. “As we continue to scale our operations across the country, this model opens doors for more inclusive innovation in education, entertainment, enterprise, and beyond. This upgrade reflects G42’s deep commitment to building AI solutions that serve India’s vibrant AI ecosystem.”
“At MBZUAI, our mission is to advance AI in ways that deliver broad, positive impact for society. NANDA marks an important milestone in bringing high-quality, open-access language technology to one of the world’s largest and most dynamic linguistic communities,” said Richard Morton, Executive Director, Institute of Foundation Models , MBZUAI. “Through our collaboration with G42 and Cerebras, we are underscoring the value of culturally aligned and inclusive AI research — work that supports underserved languages and expands access to advanced AI capabilities for hundreds of millions of Hindi speakers worldwide.”
Readers would recall that G42 had unveiled NANDA in September 2024. The announcement was made in the presence of H.H. Sheikh Khaled bin Mohammed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi, during his state visit to India. The name NANDA is inspired by one of India’s highest peaks, the Nanda Devi. Hindi is one of the most widely spoken languages in India. With over 600 million Hindi speakers and one of the world’s fastest-growing digital economies, India represents a crucial market for regional AI innovation. With over 80 percent of new internet users preferring local languages, models like NANDA can play a pivotal role in bridging digital divides.
“The upgrade of NANDA is a leap forward in making AI tools accessible for India’s vast number of users,” said Ashish Koshy, CEO of Inception. “By combining cutting-edge research with open access, NANDA is designed to serve content creators, educators, developers, and enterprises working across India’s digital landscape. This milestone reflects our continued commitment to building AI-native solutions for the Global South and accelerating AI adoption across high-impact sectors.”
The model was trained on Condor Galaxy, one of the world’s most powerful AI supercomputers for training and inferencing built by G42 and Cerebras.
NANDA 87B is now available as an open-weight model on MBZUAI Hugging Face page, enabling creators, developers, and enterprises to explore and build upon its capabilities.