The Technology Innovation Institute (TII), the applied research arm of Abu Dhabi’s Advanced Technology Research Council (ATRC), has introduced three artificial intelligence models focused on Emirati Arabic, multilingual speech recognition and Arabic optical character recognition.
The models — Falcon-Emirati, Falcon-ASR and Falcon-OCR-Arabic — extend TII’s Falcon AI portfolio across text, speech and visual content, with a particular focus on Arabic language capabilities. Falcon-Emirati is a 7-billion-parameter language model developed specifically for Emirati Arabic. It is designed to handle vocabulary, expressions, idioms and cultural references associated with the UAE dialect.
The model achieved a score of 84.83% on Alyah, a native Emirati Arabic benchmark covering everyday language, figurative expressions, heritage knowledge and poetry. According to TII, it outperformed the Arabic and multilingual open-source models included in its evaluation. H.E. Faisal Al Bannai, Adviser to the UAE President and Secretary General of ATRC, said developing AI that reflects local language and culture is an important part of the UAE’s approach to AI.
“Our language belongs in the future we are building. With Falcon-Emirati, we are putting our knowledge and expertise behind that belief,” he said. “There is real pride in building technology that understands our people. And there is real freedom in having the capability here in the UAE to develop it further, set our own priorities and turn our ambitions into something people can use.”
Building AI for spoken Emirati Arabic
The launch also addresses one of the broader challenges facing Arabic AI: the diversity of spoken Arabic. While Modern Standard Arabic is widely used in formal writing, education and news, everyday communication takes place through regional dialects, each with its own vocabulary, expressions and cultural references.
Emirati Arabic is similarly shaped by local usage, including expressions, idioms and references found in poetry, proverbs and everyday conversation. A system trained predominantly on Modern Standard Arabic may recognise the individual words in an Emirati sentence without necessarily understanding its intended meaning.
Dr Najwa Aaraj, Chief Executive Officer of TII, said the development of AI systems for local dialects is part of building AI capabilities that reflect how communities communicate in practice. “Sovereign AI capability must reflect the language used in daily life,” she said. “Emirati Arabic carries distinctive expressions, cultural references and ways of communicating. Falcon-Emirati and Falcon-ASR will help ensure that the next generation of AI understands not only Arabic but how Emirati communities actually speak it.”
Built on Falcon-H1-Arabic, Falcon-Emirati was developed using native Emirati content, Modern Standard Arabic material relating to Emirati culture and heritage, and synthetic data informed by Emirati-specific linguistic resources. The model is also being made available through TII’s Falcon Chat platform.
Speech recognition across six languages
Falcon-ASR extends the Falcon portfolio into speech recognition. The 1.6-billion-parameter model converts spoken Emirati Arabic, Modern Standard Arabic, English, French, Spanish and Portuguese into written text. Despite its relatively compact size, TII says Falcon-ASR achieved leading accuracy across its public Arabic and internal Emirati speech benchmarks, including outperforming larger systems evaluated by the institute. On Emirati speech, TII said the model surpassed the performance of a 30-billion-parameter multimodal model.
The model also records the timing of individual words within an audio file, a capability that can support applications such as subtitling, searchable recordings, meeting and interview transcription and accessibility services. The multilingual support is particularly relevant to the UAE, where Arabic and English coexist with a wide range of other languages spoken by residents and international communities.
Bringing Arabic AI to visual data
The third model, Falcon-OCR-Arabic, addresses another area of Arabic AI: extracting information from visual documents. The lightweight model is designed to recognise Arabic text and structured information in images and documents, including tables, mathematical formulas and different document sections.
Potential applications include document digitisation, searchable archives and automated information extraction, where organisations need to convert information contained in physical or digital documents into machine-readable data. The addition of OCR capabilities broadens the Falcon portfolio beyond conventional text and speech interactions, reflecting the growing demand for AI systems capable of working across multiple forms of data.
“These capabilities empower developers, researchers and communities across the region to build on Emirati language and culture, while expanding what is possible for Arabic AI more broadly,” said Dr Hakim Hacid, Chief Researcher of TII’s Artificial Intelligence and Digital Research Center. “Across Falcon-Emirati, Falcon-ASR and Falcon-OCR-Arabic, we have applied targeted specialization to language, speech and document understanding, addressing areas where Arabic AI still has significant room to advance.”
Together, the three models represent another step in TII’s development of sovereign AI capabilities, with the institute focusing on language and multimodal technologies designed around the needs of Arabic-speaking communities and the UAE’s multilingual environment.
