Photo Credit: WAM
ABU DHABI – The Technology Innovation Institute (TII), the applied research arm of Abu Dhabi’s Advanced Technology Research Council (ATRC), has launched three artificial intelligence models designed to strengthen the representation of Emirati Arabic and expand Arabic AI capabilities across language, speech and document processing.
The new models include Falcon-Emirati, a 7-billion-parameter language model developed specifically for Emirati Arabic; Falcon-ASR, a 1.6-billion-parameter automatic speech recognition model; and Falcon-OCR-Arabic, an optical character recognition model designed to extract Arabic text and structured information from images and documents.
Falcon-Emirati has been developed to understand the vocabulary, expressions, idioms and cultural context associated with the UAE’s local dialect, enabling more natural interactions between Emirati users and AI systems.
Faisal Al Bannai, Advisor to the UAE President for Strategic Research and Advanced Technology Affairs and Secretary-General of ATRC, said the development reflects the UAE’s ambition to build technology that understands the country’s people, language and priorities.
He highlighted the importance of developing sovereign AI capabilities within the UAE, allowing the country to advance technology according to its own strategic objectives while creating tools that can be used by communities, researchers and developers.
Falcon-Emirati was built on Falcon-H1-Arabic using native Emirati content, Modern Standard Arabic material related to Emirati heritage and culture, and synthetic data informed by linguistic resources specific to the Emirati dialect.
According to TII, the model achieved a score of 84.83 percent on Alyah, a benchmark for native Emirati Arabic covering everyday communication, figurative expressions, heritage knowledge and poetry. It outperformed the Arabic and multilingual open-source models included in the evaluation.
Falcon-Emirati will also be made available through the Falcon Chat platform, giving users access to AI interactions in the UAE’s local dialect.
The newly launched Falcon-ASR expands TII’s capabilities into speech recognition. The model can convert spoken Emirati Arabic, Modern Standard Arabic, English, French, Spanish and Portuguese into written text.
Despite having 1.6 billion parameters, Falcon-ASR recorded strong results across public Arabic and internal Emirati speech benchmarks, according to TII. On Emirati speech tests, the institute said the model outperformed a 30-billion-parameter multimodal system included in the evaluation.
The system also provides word-level timestamps, allowing users to identify when individual words are spoken in an audio recording. This capability can support meeting and interview transcription, subtitles, searchable recordings, accessibility solutions, multilingual customer service and voice-enabled digital platforms.
Dr. Najwa Aaraj, Chief Executive Officer of TII, said sovereign AI systems should reflect the languages people use in their everyday lives, noting that Emirati Arabic contains distinctive expressions, cultural references and communication styles.
She said Falcon-Emirati and Falcon-ASR would help future AI systems understand not only Arabic as a language but also how Emirati communities communicate.
Meanwhile, Falcon-OCR-Arabic extends the Falcon AI family into optical character recognition and document understanding.
The lightweight model can extract Arabic text and structured information from images and documents, including tables, mathematical formulas and different document sections. Potential applications include document digitisation, searchable digital archives and automated information extraction.
Dr. Hakim Hacid, Chief Researcher at TII’s Artificial Intelligence and Digital Research Centre, said the three models give developers, researchers and communities across the region additional tools to build applications around Emirati language and culture while contributing to broader progress in Arabic AI.
He added that the models apply specialised approaches across language, speech and document understanding, addressing areas where Arabic artificial intelligence still has significant potential for development.
The launch further strengthens TII’s sovereign AI portfolio and expands the Falcon ecosystem across text, speech and visual document processing, with a stronger focus on the languages and communities the technology is designed to serve.
