Cohere Transcribe Arabic ASR Model Released

Cohere has released an open-source Arabic speech-to-text model that shows superior performance in multi-dialect and code-switched Arabic and English audio transcription.

Published Jul 12, 2026, 5:57 PMUpdated Jul 12, 2026, 5:57 PM

What happened

Cohere released the open-source model Cohere Transcribe Arabic, offering superior accuracy across Arabic dialects and code-switched Arabic-English speech.

[1]

Why it matters

The model addresses a significant gap in the ASR field by improving transcription quality for various Arabic dialects and bilingual contexts, which is crucial for business and development settings.

[1]

Who is affected

Developers and businesses requiring Arabic ASR capabilities, especially in environments involving dialect variation and bilingual interactions, will benefit from this release.

[1]

Risks / uncertainty

The model requires a language tag for optimal code-switching performance, and it lacks features like timestamps and speaker diarization, which may limit its applicability in certain cases.

[1]