Hugging Face publishes Falcon ASR, a 1.6B-parameter Arabic speech recognition model with Emirati focus

Hugging Face has published Falcon ASR, a 1.6 billion parameter speech recognition model for Arabic developed at the Technology Innovation Institute (TII) in Abu Dhabi.

The model emphasizes the Emirati dialect while also supporting English, French, Spanish, and Portuguese, all using the same model weights without requiring a language flag.

In evaluations across six Arabic test sets, Falcon ASR achieved an average word error rate (WER) of 20.92%, compared to the best previously published result of 23.17% in the referenced leaderboard snapshot dated September 30, 2026.

On an internal Emirati evaluation, Falcon ASR recorded a WER of 22.73% and a character error rate (CER) of 10.19%, the lowest among the systems compared, and 4.07 percentage points better in WER than the next best result (Qwen3-Omni).

The model was trained on Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English, and was exposed to background noise, overlapping speech, music, room reverberation, telephony effects, and variations in speed and pitch to improve robustness.

Falcon ASR includes support for word-level timestamps, linking each transcribed word to its position in the audio, and builds on prior work called Falcon3-Audio.

A Hugging Face demo space is available for users to try the model, with API access and native applications planned.