Dike

Research

We study how MENA legal systems actually cite, amend, and codify law, and we publish the citation graphs, benchmarks, and datasets behind that work so AI models grounding in this region can anchor to verifiable primary authority instead of guessing.

Featured Research
Dike Model: Legal Document Typing That Knows When to Ask a Human
Model Report•September 2026•7 min read•EG, SA, AE, GCC

Dike Model: Legal Document Typing That Knows When to Ask a Human

A compact 278M-parameter model that types legal documents across five jurisdictions in real time, matches a model roughly 100 times its size where it answers, and sends everything else to review.

Model parameters: 278MDocuments per second: 845/s
Read Full Report
More Research
🤗Hugging Face Hub
Released under CC-BY-4.0 for open AI research
Early Access & Sandbox Keys

Ready to build grounded legal AI?

We are onboarding developers, legal-tech builders, and enterprise teams in cohorts. Apply on our join portal to specify your jurisdiction requirements and receive priority sandbox credentials.