TII Releases Falcon-ASR: 1.6B Arabic Speech Model Hits 20.92% WER
TII drops Falcon-ASR, a 1.6B Arabic speech model. 20.92% WER on 6 test sets — but no weights, no API. Here's what the benchmark means.
Why it matters
TII released a specialized Arabic speech-recognition model with reported WER performance on multiple dialects, but practitioners cannot yet access weights or run inference. The capability claim is specific; the deployment path is blocked.
The key facts
10 to knowModel: Falcon-ASR, 1.6B parameters
Task: Arabic automatic speech recognition with Emirati focus
Reported metric: 20.92% average WER across 6 Arabic test sets
Status: Weights NOT yet available; API NOT yet available
Source: TII (Technology Innovation Institute, UAE)
Publication: MarkTechPost rewrite of TII announcement
Model: Falcon-ASR, 1.6B parameters, Arabic with Emirati focus
Reported WER: 20.92% average across 6 Arabic test sets
Benchmark context: no comparison to other Arabic ASR systems or multilingual baselines provided
Vendor: TII (Technology Innovation Institute)
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: TII has released Falcon-ASR, a 1.6B speech recognition model for Arabic with an Emirati focus. It reports 20.92% average WER on 6 Arabic test sets, but weights and an API are not yet available.