Falcon-Emirati: New UAE-built AI model understands how Emiratis actually speak
Abu Dhabi’s Technology Innovation Institute has launched three new AI models that can understand Emirati Arabic, transcribe six languages and read Arabic documents
07 October, 2026
TT
16
Abu Dhabi’s Technology Innovation Institute (TII) has launched a new AI model built to understand Emirati Arabic, including the expressions, idioms and cultural references used in everyday conversation.
Falcon-Emirati, a 7-billion-parameter language model, was unveiled on Wednesday alongside two other AI models focused on speech recognition and Arabic documents.
The three tools can understand the UAE dialect, turn speech in six languages into text and extract Arabic information from images and documents.
TII is the applied research arm of Abu Dhabi’s Advanced Technology Research Council (ATRC).
Falcon-Emirati was developed specifically around the way Arabic is spoken in the UAE, rather than relying only on Modern Standard Arabic.
Faisal Al Bannai, Advisor to the UAE President for Strategic Research and Advanced Technology Affairs and Secretary-General of ATRC, said the development was also about building AI capability within the country.
“Our language belongs in the future we are building. With Falcon-Emirati, we are putting our knowledge and expertise behind that belief. 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.”
AI built for Emirati Arabic
Falcon-Emirati is based on TII’s Falcon-H1-Arabic model.
Researchers trained it using native Emirati material, Modern Standard Arabic content covering UAE culture and heritage, and synthetic data developed with Emirati-specific linguistic resources.
TII said the model scored 84.83 per cent on Alyah, a benchmark that tests native Emirati Arabic across everyday language, figurative expressions, heritage knowledge and poetry.
It outperformed the Arabic and multilingual open-source models included in TII’s evaluation, according to the institute.
Falcon-Emirati will be available through the Falcon Chat platform.
Dr Najwa Aaraj, CEO of TII, said local language use needed to be reflected in the development of sovereign AI.
“Sovereign AI capability must reflect the language used in daily life. 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.”
Six languages from speech to text
The second model, Falcon-ASR, converts spoken language into written text.
It supports Emirati Arabic and Modern Standard Arabic as well as English, French, Spanish and Portuguese.
Possible uses include meeting and interview transcription, subtitles, multilingual customer service, accessibility tools and other voice-based services.
The model has 1.6 billion parameters. TII said it performed strongly across public Arabic benchmarks and its own tests of Emirati speech.
On the Emirati speech benchmark, it also beat a 30-billion-parameter multimodal model included in the comparison.
Falcon-ASR can also timestamp individual words within an audio file, which could make lengthy recordings easier to search and use.
Reading Arabic documents
Falcon-OCR-Arabic, the third model announced on Wednesday, is designed to read Arabic text and structured information contained in images and documents.
It can identify content including tables, mathematical formulas and separate document sections.
The technology could be used to digitise documents and archives or extract information automatically from large numbers of records.
Dr Hakim Hacid, Chief Researcher at TII’s Artificial Intelligence and Digital Research Centre, said the three models targeted areas where Arabic AI still had scope to improve.
“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. Across Falcon-Emirati, Falcon-ASR and Falcon-OCR-Arabic, we have applied targeted specialisation to language, speech and document understanding, addressing areas where Arabic AI still has significant room to advance.”
UAE rolls out more homegrown AI
The launches add to a string of new AI models developed in the UAE this year.
In September, Mohamed bin Zayed University of Artificial Intelligence’s Institute of Foundation Models launched K2 Horizon, a family of six foundation models ranging from 0.9 billion to 375 billion parameters.
The university released the models with their weights, code, training data and methodology. Smaller versions were developed to work on devices such as phones and smart glasses, while the larger models target more demanding uses.
In July, Abu Dhabi-based Inception42 launched Seraj, an Arabic-focused enterprise AI model developed with Microsoft.
Built on GPT-4.1, Seraj was designed for Arabic-language use across government and business, including different dialects and regional context.
G42 company Inception also launched InceptionClaw in May, an AI assistant aimed at government and business users.
The system can work across tools including email, calendars and project-management platforms to prepare briefs, highlight tasks and surface information requiring attention.
TII, meanwhile, launched Falcon Perception in May.
The 600-million-parameter multimodal model combines visual and language capabilities, allowing it to analyse and interpret information contained in images.





















