Egyptian startup releases three open-source Arabic models

Alexandria-based startup Racore XYZ has launched three open-source models for Arabic business workflows: VEJI-V2, Wasl-1, and Siyaq-1. The first model, VEJI-V2, is a decision-support model with about 3.28 million trainable parameters, achieving an accuracy of 82.95% on a 522-question test subset. The second model, Wasl-1, targets cash-on-delivery orders in Egypt, while the third model, Siyaq-1, is designed to convert event streams into structured data. Racore focuses on small footprint models suitable for companies that cannot afford powerful AI servers.