JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0
Falcon ASR
Speech-to-text model from Abu Dhabi's Technology Innovation Institute for Arabic (including Emirati) and four other languages
About Falcon ASR
Arabic speakers in the Gulf are the focus of Falcon ASR, a 1.6 billion parameter speech-to-text model developed by the Technology Innovation Institute (TII) in Abu Dhabi, announced on 7 October 2026.
One model, five languages
One set of weights handles Arabic, with extra attention paid to the Emirati dialect, together with English, French, Spanish and Portuguese. Callers do not need to set a language flag, which simplifies call-centre and voice-assistant deployments that serve mixed audiences.
Reported accuracy
Word error rate is the share of words transcribed wrongly, so lower figures are better. TII reports its own evaluations. Averaged over six Arabic test sets, using the protocol of the Open Universal Arabic ASR Leaderboard, its word error rate is 20.92%, against a best published figure of 23.17% in the snapshot it used. On an internal Emirati set the figure is 22.73%, and TII says that beats Qwen3-Omni, the strongest system it compared against, by 4.07 points. For English, it reports a mean of 5.74% over seven sets taken from the Hugging Face Open ASR Leaderboard.
Open questions for buyers
These are the developer's own runs, and independent testing on real call audio would be needed to confirm them. A public demo is linked from the post, which also says that an API and native apps are planned. It does not state a licence or give a weights link, so a buyer cannot yet host the model themselves or know its terms. No price is stated.
TrustList summary
Falcon ASR is a 1.6 billion parameter speech recognition model made by Abu Dhabi's Technology Innovation Institute. The developer announced it on 7 October 2026. A single model handles Arabic, giving particular attention to the Emirati dialect, plus English, French, Spanish and Portuguese, so callers do not pick a language.
Main features
- Speech-to-text: turns spoken audio into text in Arabic and four other languages using one model.
- Emirati Arabic focus: gives particular attention to the Emirati dialect, and the developer reports its own result on an internal Emirati test set.
- Developer benchmarks: in its own runs the developer reports a 20.92% word error rate on six Arabic test sets and 5.74% on seven English sets.
- Demo and access: the announcement links a public demo, and says API access and native apps are planned.
Key facts
- Developer: Technology Innovation Institute
- Based in: Abu Dhabi
- Announced: 7 October 2026
- Parameters: 1.6 billion
Summary written by TrustList from the business's own published information, Nov 2026.
Benchmarks & AI stats
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Falcon ASR launches
Request a demo or quote from Falcon ASR
Reviews
Write the first review of Falcon ASR
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about Falcon ASR.
Falcon ASR alternatives
Similar listings buyers compare against this one.
Microsoft AI's high-volume, low-latency variant of MAI-Voice-2.1 for speech in 23 languages, aimed at voice agents
Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure
Aleph Alpha's open-weight English-German mixture-of-experts model for self-hosting, released under Apache 2.0 in October 2026
Ai2's open-weight 8B model that writes cited scientific reports from a research question and retrieved literature
Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection
Black Forest Labs' image generation and editing model with bounding-box layouts, multiple reference images and 4K output
Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.