Kuwain
A 1.5B-parameter Arabic SLM built by language injection, the base layer for our downstream models.
We understand deeply. We build from the source. And we run with confidence.
Arabic foundation models from Kawn Lab: language, vision, speech, embedding and specialized NLP.
A 1.5B-parameter Arabic SLM built by language injection, the base layer for our downstream models.
A family of Arabic dialect translation models covering dialect-to-dialect and dialect-to-MSA translation.
Bidirectional Arabic–English translation built on Kuwain, for general and specialized domains.
Arabic diacritization built on Kuwain, restoring full Tashkeel for MSA and classical Arabic.
Arabic document intelligence: scans to structured Markdown
Lightweight general-purpose Arabic embeddings for semantic search, retrieval, clustering, and RAG.
Arabic embeddings optimized for retrieval across the Quran, Hadith, Fiqh, and Fatwa collections.
Domain-specific Arabic embeddings for healthcare applications.
Acoustic model for Quranic recitation: ASR and Tajweed-aware mispronunciation detection.
High-quality Najdi Arabic text-to-speech for conversational AI and accessibility applications.
Hierarchical classification that routes Fatwa and Fiqh questions to the correct legal category.
Information extraction that identifies verbatim Quran, Hadith and book quotations.
Tell us the problem. We'll tell you honestly whether AI is the right answer.