Cohere Transcribe Arabic
Cohere Transcribe Arabic adapts the two-billion-parameter Transcribe model for Arabic speech. It supports regional dialects and Arabic-English code-switching, with Apache 2.0 weights and an evaluation API.
Capabilities and limits
The model is intended to handle dialect, accent and recording variation. It does not provide timestamps or automatic speaker identification. Cohere documents Model Vault for production inference without the evaluation API's rate limits.Cohere Transcribe Arabic documentation
When this model fits
Practical guidance. This is the more focused comparison for calls that move between Arabic and English. Review local dialects, company names and specialized vocabulary using real examples with permission. Decide whether the desired transcript should preserve spoken wording or be edited later into formal text. Keep transcription and editorial rewriting as separate steps so reviewers can see where the wording changed.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable |
| Model identifiers | cohere-transcribe-arabic-07-2026 |
| Maximum API file size | 25 MBCohere Transcribe Arabic documentation |
| Language scope | Arabic and EnglishCohere Transcribe Arabic documentation |
| Weight license | Apache 2.0Cohere Transcribe Arabic documentation |
Related models
- Cohere Transcribe
Cohere · Transcribe
Two-billion-parameter speech recognition model for 14 languages, with Apache 2.0 weights.
Launch date not confirmedAPI · Downloadable