Community-tested models

AI models tested for Arabic

Does it actually work for Saudi, Najdi, Hijazi or Gulf Arabic? Benchmarks rarely say. The community tests these models on real speech and text — and reports back.

All ASR TTS Translation LLM Embedding Vision Other2 models

All languages ar en es fr multi pt zh

All dialects Egyptian Gulf Levantine MSA Najdi Saudi

Arabic support:All Full Partial None

All licenses Apache 2.0 AudarAI Community License v1.0 CC-BY-NC-4.0 Gemma Terms of Use Llama 4 Community License MIT Proprietary (API only)TII Falcon LLM License

Audar ASR V1 Turbo

Audar AI Labs

ASR

No evaluations yet

Arabic: fullMSAGulfEgyptianLevantinearenmulti

2.35B Arabic-first speech recognition trained on 300k+ hours, naming Gulf, Egyptian, Levantine and Maghrebi coverage plus Arabic-English code-switching. Its card claims the top place on the Open Universal Arabic ASR Leaderboard — a leaderboard run by Elm, a Saudi company. Custom AudarAI Community Licence, not open source.

Hugging Face

Munsit

CNTXT AI

ASR

No evaluations yet

Arabic: fullMSAGulfEgyptianLevantinear

Arabic ASR trained with weak supervision on 30K+ hours; the paper claims best-in-class accuracy across 18 dialects. No weights are published — there is no CNTXT organisation or Munsit repository on Hugging Face — and the model is commercial API-only, so the dialect claims cannot be independently checked.