As of Oct 8, Mistral published LIDstral-Arabic on Hugging Face under the Apache 2.0 license.
Timeline · newest first
Official
Mistral releases LIDstral-Arabic, an open language and dialect identification model for Arabic
Hugging Face @mistralai published this on Oct 8, 2026 (source: huggingface.co).
Mistral published LIDstral-Arabic on Hugging Face under the Apache 2.0 license. The fastText-based classifier identifies Modern Standard Arabic and Maghrebi, Gulf, Nile Valley and Levantine Arabic dialects, along with Berber, Turkic, Indo-Iranian and African languages often found alongside Arabic text.