MISRAJAI
What we believe
Arabic-first by design. Built for the language, not translated into it.
Arabic-first by design. Built for the language, not translated into it.
Misraj AI builds Arabic AI from the ground up, turning original research and proprietary technology into secure, intelligent solutions for enterprises, governments, and the region.
2 trillion
Tokens for Arabic-aware models.
35B+
Open-source tokens for Arabic research.
10+
Papers that test what we build and learn.
100k+
Users and beneficiaries of Misraj products.
30+
Solutions, products and models in real use.
1st
Manuscript reading, NAKBA NLP 2026.
Turn Arabic documents into structured data you can search, connect to your systems, and build on.
Platforms you build on, and systems delivered ready to run.
Smart conversation, inside your organisation's walls
A secure enterprise chat experience, connected to your data sources and documents, giving your teams more accurate answers without sending sensitive knowledge to public tools.
Designed for trust
Every model in the catalogue runs inside your boundary, on your hardware or in-Kingdom, designed to SDAIA guidance, NCA ECC and PDPL.
Every model behind our products comes out of our own lab, and the work ships in the open: papers, open weights, and the benchmarks nobody else had built.
Nakba-NLP 2026 @ LREC 2026
arXiv (2025), LaTell 2026
arXiv (2025), LaTell 2026
Page 1 of 3
Models built for the rules and the workload of every sector we serve in the Gulf.
Page 1 of 2
It means the Arabic is trained in rather than translated on. The pre-training corpus is built for Arabic language-model pre-training rather than scraped as an afterthought, the models are trained on it, and they are measured on Arabic evaluations we had to build ourselves because the existing ones were too narrow to settle the question. An interface localised after the product was designed is a different thing, and this site tries never to use the phrase for that.
Tell us the problem. We'll tell you honestly whether AI is the right answer.
