Baseer by Misraj AI
Riyadh-built Arabic document OCR producing Markdown with tables; thin docs
Visit website →About
Baseer is an Arabic document-OCR model from Misraj AI, a Riyadh company, described in a September 2025 paper and a December 2025 launch announcement. It is a vision-language model fine-tuned for Arabic pages that converts a scanned or photographed page into Markdown, preserving headings, paragraphs and tables and handling mixed Arabic and English text. On the company's own Misraj-DocOCR benchmark the paper reports a word error rate of 0.25, and the launch material positions it against general-purpose models for Arabic documents.
What is not published matters as much: there is no independent benchmark result, no documented handwriting capability, no pricing page, no statement of where processing happens, and no information on ID or invoice templates. We list it because a Saudi-built Arabic document model is exactly what this guide's readers want to exist, and because the gap between its claims and its documentation is something a buyer should ask Misraj to close before a pilot. Pricing on request.
AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.
Why we picked it
A Saudi company training an Arabic-first document model is the development this whole guide is waiting for. We list Baseer on its paper and launch announcement, and say exactly what is missing, so Misraj knows what documentation would earn a stronger listing and readers know what to request.