---
title: "Introducing Falcon ASR"
description: "The Technology Innovation Institute in Abu Dhabi released Falcon-ASR, a 1.6 billion parameter speech recognition model focused on Arabic, especially the Emirati dialect, while also supporting English, French, Spanish, and Portuguese. On the Open Universal Arabic ASR Leaderboard, Falcon-ASR achieved a 20.92% average word error rate across six test sets, outperforming the best published average of 23.17% in the reference snapshot. The model recorded a 22.73% word error rate and 10.19% character error rate on an internal Emirati evaluation, 4.07% better in word error rate than the next best system compared. Training data included background noise, overlapping speech, music, reverberation, telephony effects, speed, and pitch variations to handle diverse recording conditions. Falcon-ASR builds on the Falcon3-Audio architecture and provides word-level timestamps linking transcribed words to audio positions."
---

# Introducing Falcon ASR

[Hugging Face](https://yomu.fyi/company/hugging-face) · Abdul Muneer, Rishabh Saraf, Ramesh Gundluru, Shamsa Hamad, Lepauloux, Hakim Hacid · Oct 7, 2026

**Type:** Announcement

## Summary

The Technology Innovation Institute in Abu Dhabi released Falcon-ASR, a 1.6 billion parameter speech recognition model focused on Arabic, especially the Emirati dialect, while also supporting English, French, Spanish, and Portuguese. On the Open Universal Arabic ASR Leaderboard, Falcon-ASR achieved a 20.92% average word error rate across six test sets, outperforming the best published average of 23.17% in the reference snapshot. The model recorded a 22.73% word error rate and 10.19% character error rate on an internal Emirati evaluation, 4.07% better in word error rate than the next best system compared. Training data included background noise, overlapping speech, music, reverberation, telephony effects, speed, and pitch variations to handle diverse recording conditions. Falcon-ASR builds on the Falcon3-Audio architecture and provides word-level timestamps linking transcribed words to audio positions.

## Context

Arabic speech recognition faces challenges from regional dialect variation, differences between Modern Standard Arabic and dialectal forms, and a shortage of transcribed resources for dialects. Existing systems often struggle with conversational Emirati speech or recordings made over phone lines. The Open Universal Arabic ASR Leaderboard maintained by the ELM Research Center ranks systems by average word error rate across six test sets.

## Approach / What changed

Training Falcon-ASR as a 1.6 billion parameter speech recognition model that uses a single set of weights to support Arabic (including the Emirati dialect), English, French, Spanish, and Portuguese without requiring a language flag. The training data includes Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English, augmented with background noise, overlapping speech, music, reverberation, telephony effects, speed and pitch variations. The model builds on the Falcon3-Audio architecture described in the paper Competitive Audio-Language Models with Data-Efficient Single-Stage Training on Public Data.

## Takeaways

- Falcon-ASR achieves a 20.92% average word error rate across six Arabic test sets on the Open Universal Arabic ASR Leaderboard, compared to the best published 23.17%. Its internal Emirati evaluation shows 22.73% WER, 4.07% lower than the next best system tested.
- A single set of weights handles Arabic, English, French, Spanish, and Portuguese without a language flag. Training data includes Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English.
- Training augmentation covered background noise, overlapping speech, music, reverberation, telephony effects, and speed/pitch variation to improve robustness across recording conditions.

## FAQ

### Does Falcon-ASR require a separate model for each language?

Falcon-ASR is a single 1.6 billion parameter model that uses the same weights for all five languages. It does not need a language flag; it outputs a transcript in the language spoken.

### Does Falcon-ASR support word-level timestamps?

Yes, the model provides word-level timestamps that link each transcribed word to its position in the audio recording.

### How can I use Falcon-ASR?

Currently a Hugging Face demo space is available for trying the model. API access and native applications are planned but not yet released.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Arabic](https://yomu.fyi/topic/arabic), [ASR](https://yomu.fyi/topic/asr), [Open Source](https://yomu.fyi/topic/open-source)

- Source: [Hugging Face](https://huggingface.co/blog/tiiuae/falcon-asr)
- Source URL: https://huggingface.co/blog/tiiuae/falcon-asr
- Ingested by Yomu: 2026-10-08T15:00:13.375Z

[Read original post](https://huggingface.co/blog/tiiuae/falcon-asr)