Concerns that Arabic might be left behind in the fast-developing AI sector are starting to evaporate with the introduction of the Falcon Arabic language model, created in Abu Dhabi.
The model was unveiled on Wednesday by the Technology Innovation Institute (TII), an Abu Dhabi government-backed research centre which first introduced its Falcon large language model back in 2023.
Faisal Al Bannai, adviser to the UAE President for Strategic Research and Advanced Technology Affairs, spoke about the development as a leap forward for Arabic at the UAE’s Make it in the Emirates event.
“We’re proud to finally bring Arabic to Falcon, and prouder still that the best-performing large language model in the Arab world was built in the UAE,” he said.
According to TII, Falcon Arabic is trained on a native (non-translated) Arabic data set that covers both Modern Standard Arabic and regional dialects.
“It captures the full linguistic diversity of the Arab world,” said TII.
The research centre also said that so far the model outperforms other Arabic language models.
Large language models are complex systems designed to be trained on large amounts of text and data that help AI implementations identify patterns, come to conclusions and even understand nuances. In short, the models can make or break the user experience with AI.
Although Arabic is spoken by about 400 million people worldwide, it was not initially a focus during the initial growth of AI and large language models, with English the most prevalent.
The complexity and diversified Arabic dialects, coupled with various language nuances, posed a challenge for engineers and programmers trying to perfect machine learning technologies.
In recent years, the UAE has sought to bolster Arabic’s presence in the AI race.
In 2023, Jais, an open-source bilingual Arabic-English model, was introduced by G42, Mohammed bin Zayed University of Artificial Intelligence and Silicon Valley-based Cerebras Systems.
Later that year, Jais Climate, the world’s first bilingual large language model dedicated to climate intelligence was also announced.
In addition to Falcon Arabic, TII also announced on Wednesday the release of its Falcon H1 model, which it says “outperforms comparable offerings from Meta's LLaMA and Alibaba's Qwen, enabling real-world AI on everyday devices and in resource-limited settings".
The research centre explained that efficiency was at the core of Falcon H1 development.
“This fundamentally shifts what’s possible at the smallest scale, enabling powerful AI on edge devices where privacy, efficiency, and low latency are critical,” said Hakim Hacid, chief researcher at the TII AI and digital science research centre.
“It demonstrates how new architectures can unlock new opportunities in AI training while showcasing the potential of ultra-compact models.”
