🔍 Read the full analysis: Getting Started With Falcon ASR For AI Transcription 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
Abu Dhabi’s Technology Innovation Institute has introduced Falcon-ASR, a 1.6-billion-parameter speech recognition model focused on Arabic, including the Emirati dialect. TII reports a 20.92% average word error rate across six Arabic test sets and a 22.73% rate in an internal Emirati evaluation; users can try a Hugging Face demo, while API access and native apps are planned.
The Technology Innovation Institute (TII) in Abu Dhabi has introduced Falcon-ASR, a 1.6-billion-parameter speech recognition model designed for Arabic, with particular attention to the Emirati dialect, as described in the original analysis. TII says the model supports five languages and reports benchmark results for Arabic and English; the model is available to try in a Hugging Face demo, while API access and native applications are planned.
TII reports an average word error rate (WER) of 20.92% across six Arabic test sets used by the Open Universal Arabic ASR Leaderboard. The institute says the best published average in the leaderboard snapshot it checked on 30 September 2026 was 23.17%, a difference of 2.25 percentage points. The leaderboard, maintained by the ELM Research Center, gives equal weight to the six test sets; a lower WER means fewer word-level transcription errors. TII says it followed the leaderboard protocol and used its pinned manifests.
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. The institute says these were the lowest scores among the systems it compared, and that the next-best WER was Qwen3-Omni’s, 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.
The model returns word-level timestamps and, according to TII, uses the same weights for Arabic, English, French, Spanish and Portuguese without requiring a language flag. The institute says its training included Emirati, Modern Standard Arabic, other Gulf and Arabic dialects, and English, alongside audio conditions such as background noise, overlapping speech, music and telephony effects. These details and performance figures are claims reported by the model’s developer.
Testing Arabic Beyond Formal Speech
Falcon-ASR’s reported Emirati results address a practical challenge for speech recognition: performance on everyday dialect speech may differ from performance on formal Arabic, and dialectal training material is less available than material for Modern Standard Arabic. A model that handles casual conversations, calls or language switching could be useful to developers building transcription tools for meetings and recordings.
The benchmark figures are a starting point, not a guarantee for every use. TII’s scores come from specified evaluation sets, and they do not establish how Falcon-ASR will perform for every speaker, dialect, microphone or noisy setting. Word-level timestamps may help users find passages in longer recordings, but practical value will depend on results with the audio and workflows people actually use.
Arabic speech recognition software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How TII Benchmarked Falcon-ASR
The Arabic comparison uses an equal-weight average across six test sets on the Open Universal Arabic ASR Leaderboard. TII’s comparison is tied to the snapshot it checked on 30 September 2026, rather than a live ranking. The supplied announcement does not give Falcon-ASR’s individual scores on each set, so readers cannot see from that information which test sets contributed most to the average.
TII describes the Emirati results as an internal evaluation on held-out recordings and says public coverage also includes the UAE subset of the Casablanca dataset. The institute says Falcon-ASR builds on its Falcon3-Audio work. The available material does not describe an independent replication of the new model’s results, so the reported figures should be read as developer-reported evaluations.
“Our aim is to transcribe the words people use in everyday speech, including dialectal forms and switches between languages.”
— Technology Innovation Institute
AI transcription tools for Arabic dialects
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Available Evaluation
The announcement does not provide a full breakdown by test set, dialect, speaker or recording condition. It also does not state the size and detailed composition of the internal Emirati evaluation or list every system included in that comparison. Those details would help readers judge how broadly the reported scores apply.
The Arabic leaderboard comparison reflects a snapshot checked on 30 September 2026, and later results could change the comparison. The source material describes the figures as TII’s evaluations and does not report an independent replication. Real-world performance remains unestablished for recordings that differ from the test material, including varied accents, background conditions and speaker patterns.
speech-to-text API for Emirati Arabic
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Demo Access and Planned Releases
People can try Falcon-ASR through TII’s Hugging Face Demo Space, which the institute says accepts users’ recordings for transcription. The announcement says API access and native applications are planned but gives no release dates. Until those options become available, the demo is the stated route for testing the model directly.
Further evaluation details, including per-dialect results and independent testing, would help clarify how well the reported performance carries over to different users and recording conditions. For now, the benchmark results describe performance on the identified test material, while individual users can use the demo to examine output on their own recordings.
multilingual speech recognition app
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 supports Arabic, English, French, Spanish and Portuguese, with particular attention to Arabic and Emirati speech.
What Arabic benchmark result did TII report?
TII reports an average 20.92% word error rate across six Arabic test sets on the Open Universal Arabic ASR Leaderboard. The institute compared that result with a 23.17% best published average in the snapshot it checked on 30 September 2026.
How did Falcon-ASR perform in TII’s Emirati evaluation?
TII reports 22.73% WER and 10.19% character error rate on an internal evaluation using held-out Emirati and Gulf recordings with human-validated transcripts. The institute says these were the lowest scores among the systems it compared.
How can people try the model?
TII says Falcon-ASR can be tried through its Hugging Face Demo Space. API access and native applications are planned, but the announcement does not give release dates.
Are the reported scores independently verified?
The supplied announcement presents the results as TII-reported evaluations and does not describe an independent replication. It also does not provide a complete breakdown of results by dialect, test set or recording condition.
Primary source: Hugging Face · via ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
