TII in Abu Dhabi introduced Falcon-ASR, a 1.6B-parameter speech recognition model for Arabic, focused on Emirati dialect, that also handles English, French, Spanish and Portuguese. TII reports a 20.92% average word error rate on six Arabic test sets.
TII says one set of weights covers all five languages without a language flag and returns word-level timestamps. It reports 20.92% average word error rate against 23.17% for the best published result in a leaderboard snapshot it checked on 30 September, and 22.73% on its own internal Emirati test set. TII says a demo is live and an API and native apps are planned.
Sources: TII (7 Oct) · Note: Every score is TII’s own, and the Emirati test set is TII’s internal set. CTN has not verified the scores, and the leaderboard comparison reflects a 30 September snapshot. Licence and weights availability are unconfirmed.