🔍 Read the full analysis: Introducing Falcon ASR on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
The Technology Innovation Institute in Abu Dhabi has introduced Falcon-ASR, a 1.6-billion-parameter speech recognition model focused on Arabic and the Emirati dialect. TII reports a 20.92% average word error rate across six Arabic test sets and the lowest error rates in its internal Emirati comparison; the figures have not been independently replicated in the source material.
The Technology Innovation Institute (TII) in Abu Dhabi has introduced Falcon-ASR, a 1.6-billion-parameter speech recognition model focused on Arabic, particularly the Emirati dialect, as described in the original analysis. TII reports a 20.92% average word error rate across six Arabic test sets and says the model had the lowest error rates among systems in its internal Emirati evaluation; the model is available to try in a Hugging Face demo.
TII says Falcon-ASR also transcribes English, French, Spanish and Portuguese, using the same model weights for all five languages without requiring users to specify a language flag. The model returns word-level timestamps, linking each transcribed word to its position in the audio. According to the institute, its training included Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English, along with audio conditions such as noise, overlapping speech, music, reverberation and telephony effects.
For Arabic, TII reports an equal-weight average WER of 20.92% across the six test sets used by the Open ASR Leaderboard. The institute compared that result with a 23.17% best published average in the leaderboard snapshot it checked on 30 September 2026, a difference of 2.25 percentage points. The leaderboard is maintained by the ELM Research Center; TII says it followed the protocol and used the pinned manifests. Lower WER means fewer word-level transcription errors on the evaluation material.
For Emirati speech, TII reports 22.73% WER and 10.19% character error rate in an internal evaluation using held-out Emirati and Gulf recordings with human-validated transcripts. It says these were the lowest scores among the systems it compared, and that the next-best WER, from Qwen3-Omni, was 4.07 percentage points higher. TII also reports a mean WER of 5.74% on seven public English test sets used by the Hugging Face Open ASR Leaderboard.
A Test for Emirati Everyday Speech
Speech recognition performance on spoken dialects matters to users building tools for meetings, phone calls and everyday recordings. Arabic varies across regions and settings, while transcribed training material is less available for many dialects than for Modern Standard Arabic. A system that performs well on formal speech may not handle casual Emirati conversation, Gulf accents or language switching as reliably.
Falcon-ASR’s reported Emirati evaluation puts a specific dialectal use case alongside its broader Arabic benchmark result. If the results hold up across more speakers and recording conditions, the model could be relevant to developers working on Arabic transcription services. Its word-level timestamps may also help users find particular statements in longer recordings. The reported scores, however, describe performance on defined test material; they do not establish results for every speaker, accent or application.
Arabic speech recognition software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How TII Measured Recognition Accuracy
The Arabic comparison uses six test sets and an equal-weight average of WER scores, according to TII’s account of the leaderboard protocol. Its comparison is tied to a snapshot checked on 30 September 2026, rather than a live ranking. The source material does not provide Falcon-ASR’s individual result for each test set, so readers cannot see how performance varied across them.
TII describes the Emirati assessment as an internal evaluation using held-out recordings and transcripts validated by people. It also points to the public Casablanca dataset, which includes a UAE subset. The model builds on TII’s Falcon3-Audio work, which the institute says informed its architecture and training approach. These details provide context for the reported results but do not amount to an independent evaluation.
“Our aim is to transcribe the words people use in everyday speech, including dialectal forms and switches between languages.”
— Technology Innovation Institute
multilingual speech-to-text transcription device
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Published Evaluations
The available announcement does not include a full breakdown by test set, dialect, speaker or recording condition. It also does not state the size and composition of the internal Emirati evaluation or identify every system included in that comparison. Those details would help readers judge how broadly the reported WER and character error rate apply.
The scores are reported by the model’s developer; the source material does not describe an independent replication. The Arabic comparison reflects a leaderboard snapshot checked on 30 September 2026, and later published results could change how Falcon-ASR compares. How it performs on recordings outside the evaluation material, including different accents and real-world noise, remains unestablished.
English Arabic language transcription tool
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Demo Available, Releases Undated
Users can try Falcon-ASR through TII’s Hugging Face Demo Space, which the institute says allows testing with personal recordings. TII says API access and native applications are planned, but the announcement gives no release dates.
More detailed test results, per-dialect breakdowns and independent evaluations would help clarify the model’s performance beyond the reported comparisons. Until those details or additional releases are available, the demo can show how the system handles particular recordings, while the benchmark figures remain evidence from the specified test sets.
voice recognition microphone for Arabic dialects
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Falcon-ASR?
Falcon-ASR is a 1.6-billion-parameter speech recognition model introduced by Abu Dhabi’s Technology Innovation Institute. TII says it focuses on Arabic, especially Emirati speech, and also supports English, French, Spanish and Portuguese.
What Arabic benchmark result did TII report?
TII reports an average word error rate of 20.92% across six Arabic test sets in the Open Universal Arabic ASR Leaderboard protocol. It compared that with a 23.17% best published average in a snapshot checked on 30 September 2026.
How did Falcon-ASR perform on Emirati speech?
TII reports 22.73% WER and 10.19% character error rate in its internal Emirati evaluation. The institute says these were the lowest scores among the systems it compared; independent replication is not described in the source material.
Can people try Falcon-ASR now?
Yes. TII says a Hugging Face demo is available for testing recordings. The institute has also said API access and native applications are planned, but has not announced dates for them.
Are the reported scores proof of performance in every setting?
No. The figures relate to specified test sets and an internal Emirati evaluation. TII has not published a full breakdown by speaker, dialect or recording condition, and the available source does not describe an independent replication.
Primary source: Hugging Face · via ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
