What did Abu Dhabi's Technology Innovation Institute release?

It released the Falcon-Emirati model and related tools for speech recognition and extracting Arabic text from images and documents, all trained on the Emirati Arabic dialect.

How does the speech recognition model compare with larger models?

The speech recognition model has 1.6 billion parameters, but according to TII it is more accurate than a 30-billion-parameter model.

Why has Arabic AI training lagged behind other languages?

Partly because dialects vary widely and can be incomprehensible to people who know only the formal Arabic used in news reports and textbooks, which has made some machine translations comically inaccurate.

Large Language ModelsAudio & SpeechOpen-Source AISemafor Tech18h agoOct 9, 2026, 01:00 JST

Abu Dhabi's TII releases Falcon-Emirati dialect AI models

Send on LINE

3 Key Points

What happened
Abu Dhabi's state-run Technology Innovation Institute released Falcon-Emirati models trained on the Emirati Arabic dialect, improving translation, transcription, and voice-enabled services; its 1.6 billion-parameter speech recognition model is more accurate than a 30-billion-parameter model, per TII.
Why it matters
The released models are relatively compact yet the smaller speech model outperforms a much larger one on accuracy, according to TII, suggesting capable Arabic-dialect AI may not require very large models.

WHO IT HITSBusinesses and public services in the Gulf that rely on Arabic-language transcription, translation, or voice interfaces may see fewer errors, since the models are trained on the Emirati dialect rather than only formal Arabic.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

TII is a state-run research body in Abu Dhabi, and its decision to train on the Emirati dialect addresses a specific problem: Arabic dialects vary so widely that they can be incomprehensible to speakers of the formal Arabic found in news reports and textbooks. That gap has produced machine translations that are sometimes comically inaccurate, occasionally turning innocuous words into vulgar English references. By training on local dialects, idioms, and cultural references, TII says it aims to eliminate those errors and make AI more useful in the region. The released models are relatively compact, and the speech recognition model in particular pairs a small parameter count with higher accuracy than a much larger model, according to TII.

FAQ

What did Abu Dhabi's Technology Innovation Institute release?: It released the Falcon-Emirati model and related tools for speech recognition and extracting Arabic text from images and documents, all trained on the Emirati Arabic dialect.
How does the speech recognition model compare with larger models?: The speech recognition model has 1.6 billion parameters, but according to TII it is more accurate than a 30-billion-parameter model.
Why has Arabic AI training lagged behind other languages?: Partly because dialects vary widely and can be incomprehensible to people who know only the formal Arabic used in news reports and textbooks, which has made some machine translations comically inaccurate.

Semafor TechRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.