A cutting-edge Abu Dhabi university has unveiled an Arabic AI tool designed to decode a variety of regional dialects and respond to rapid shifts in language.
Jais 3 – built by the Mohamed bin Zayed University of Artificial Intelligence, in partnership with Abu Dhabi's Inception42 – is able to field queries and carry out tasks such as transcribing conversations and creating documents and presentations.
It features Jais speech, a voice recognition model for the Arabic language and its varying dialects from across the Gulf and North Africa.
The advanced AI platform was developed through the university's Institute of Foundation Models (IFM) with the goal of keeping pace with the nuances of a language spoken by more than 400 million people around the world. The tech is also able to deal with code-switching, which refers to when a bilingual person alternates from one language to another.
"Arabic AI needs to understand the language in its everyday context, which varies by geography and across generations," said Haonan Li, senior research scientist at IFM. "For example, we need to account for the many nuances in spoken Arabic, including when speakers slip into English or French mid-sentence and back again. We built Jais 3 for Arabic as it's really used."
High-tech training
Jais Speech transcribes Modern Standard Arabic and dialects spoken across 19 Arab countries. It follows speakers as they switch between Arabic and languages such as English or French, keeping those words in their original script. It also identifies the dialect being spoken and can add Arabic vowel marks to a transcript.
Paired with the language models, Jais Speech enables Jais 3 to take spoken instructions in Arabic or English and act on them.
Mr Li said the team worked extensively to ensure strong representation of the many forms of Arabic spoken in the region and beyond.
“One of the key challenges was ensuring broad and balanced representation across regional dialects," he said. "For Gulf dialects, this involved working with a more limited pool of existing open data and addressing the tendency of open models to default to Modern Standard Arabic rather than maintaining the requested dialect when generating synthetic data.
"This required additional work in data synthesis and filtering to ensure the quality and relevance of the resulting data set."
A similar challenge was faced by the Jais 3 team for Morocco, where available data tended to be heavily weighted towards Moroccan Darija, leaving other dialects in the region less represented.
Embracing technology
Richard Morton, vice president and executive director of IFM, said the launch of Jais 3 showed the importance of investing in rapidly evolving technology.
"Jais has shown what the UAE's long-term investment in open AI research can deliver for the Arabic-speaking world," he said. "With Jais 3, we are extending that foundation into speech and agents at a time when the UAE is putting AI to work across government and society – reflecting MBZUAI's commitment to the UAE National AI Strategy 2031 and to building the research capability and national talent that will carry it forward."
IFM and Inception42 evaluated the models on an audited suite of 24 Arabic benchmarks, which they are releasing publicly alongside the models. The results showed the largest Jais 3 model leads open models on agentic tasks and matches the strongest on Arabic understanding.
Jais Speech recorded the lowest average word and character error rates among all models compared, while the seven billion and 32 billion-parameter models earned the highest average Arabic scores among open models of comparable sizes.
"Jais 3 makes Arabic AI more practical for business use," said Dr Larry Murray, senior vice president of Applied Science at Inception42. "Stronger reasoning and instruction following, combined with tool use, provide a foundation for multi-step agentic workflows. With open models across a range of sizes, organisations can match these capabilities to their computing budgets and deployment needs."