Sources & directories

Searchable map and directory of the Arabic AI ecosystem, with an offline MCP server and agent skills. A directory, not an individual AI model.

Curated list of Arabic LLMs, speech models and datasets.

About 1,500 startups at the 2025 show; 223 MENA AI startups, 15 focused on Arabic or Gulf markets.

Community page with about 57 models, including Saudi speech-to-text, voice models and Qari Arabic OCR.

Arabic educational web text, including FineWeb-Edu-Arabic.

About 12 Arabic training datasets.

Arabic voice and Arabic search leaderboards, and the Yehia-7B chat model.

About 1,100 exhibitors in 2026, including over 300 North African AI companies, many Moroccan.

Arabic training data, speech datasets and AI-generated text detection.

ARBML catalogue of Arabic NLP and speech datasets. Check each dataset for its own access and licence; dataset counts change.

Has an AI & Analytics category with 46 apps for Zid stores.

Has a MENA region filter: 91 companies, a handful of them Arab AI startups.

Public database of 921 labelled startups with CSV download; about 70 build AI.

316 startups with sector and country tags; 42 are MENA-based AI companies.

Egypt's Applied Innovation Center: Karnak 40B and 6B models and AIC-1.

ArabianGPT models and Arabic OCR data.

About 600 company profiles; mostly older cohorts, with new Saudi AI startups from 2025.

QCRI's page: about 39 models and 64 datasets, including Arabic dialect benchmarks and the Fanar models.

Searchable exhibitor list for 7–11 Dec 2026; still filling, with 41 MENA AI exhibitors so far, mostly integrators and government.

Has an AI-Driven Solutions category with 19 apps for Salla stores.

Moroccan Darija OCR, models and the Darija chatbot arena.

11 datasets for Arabic image captions, document OCR and diacritics.

Tunisian Arabic speech-to-text and Tunisian datasets.

About 44 models for Arabic search and text matching, 31 datasets and an Arabic MMMLU leaderboard.

About 60 models and 41 datasets, including NileChat, MARBERT and AraT5.

Atlas-Chat for Moroccan Darija, Nile-Chat for Egyptian Arabic, and 22 datasets.

About 293 Arabic datasets, including the Masader catalogue.

Open Hugging Face community for Arabic speech: datasets, models, benchmarks and demos for MSA and Egyptian, Levantine, Gulf and Maghrebi dialects, covering speech recognition, text-to-speech and more.

Collection of 200+ Arabic language datasets, plus a collection for each country.