Skip to content

Dataset · ClusterlabAi

Arabic text corpus with 101B words

ClusterlabAi/101_billion_arabic_words_dataset

A large Arabic text corpus for training and fine-tuning language models.

What's inside

A collection of Arabic texts gathered by the Clusterlab team, with a stated size of 101 billion words in the card. Data is maintained with weekly updates to improve quality and coverage. A cleaner version with a UUID column for traceability is in preparation; metadata lists the size as 10M<n<100M, and the language as Arabic.

What you can do with it

  • Training Arabic LLMs
  • Continued pretraining in Arabic
  • Evaluating language coverage
Tasks
Text generation
Languages
Arabic
Size
10M–100M rows
Downloads this month
3.1K
Likes
73

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ClusterlabAi/101_billion_arabic_words_dataset")