Jais 2: A Family of Arabic-Centric Open Large Language Models
Source
arxiv.org
Author
Mohamed Anwar, Abed Alhakim Freihat, George Ibrahim, Mostafa Awad, Abdelrahman Sadallah, Gurpreet Gosal, Gokulakrishnan Ramakrishnan, Sarath Chandran, Biswajit Mishra, Rituraj Joshi, Ahmed Frikha, Etienne Goffinet, Abhishek Maiti, Ali El Filali, Sarah AlBarri, Samujjwal Ghosh, Rahul Pal, Parvez Mullah, Awantika Shukla, Sajid siddiki, Samta Kamboj, Onkar Pandit, Sunil Kumar Sahu, AbdelRahman Elbadawy, Amr Mohamed, Ahmad Chamma, Evan Dufraisse, Abdelaziz Bounhar, Dani Bouch, Hadi Abdine, Guokan Shang, Fajri Koto, Yuxia Wang, Zhuohan Xie, Ali Mekky, Rania Elbadry, Sarfraz Ahmad, Momina Ahsan, Omar El Herraoui, Daniil Orel, Hasan Iqbal, Kareem Elzeky, Mervat Abassy, Kareem Elozeiri, Saadeldine Eletter, Farah Atif, Nurdaulet Mukhituly, Haonan Li, Xudong Han, Aaryamonvikram Singh, Zainul Abedien Ahmed Quraishi, Neha Sengupta, Larry Murray, Avraham Sheinin, Joel Hestness, Natalia Vassilieva, Hector Xuguang Ren, Zhengzhong Liu, Michalis Vazirgiannis, Preslav Nakov
Date
Context
Jais 2 offers open weightsA model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.Full definition → language models centered on Arabic, with 70-billion-parameter and eight-billion-parameter versions. The authors describe an Arabic-focused vocabulary and training approach intended to make efficient use of computation while supporting both Arabic and English.
The report evaluates more than general language tasks: it includes culturally groundingTying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.Full definition → Arabic tests involving poetry, religion and cuisine. The authors report leading results among the open models they evaluated on two named Arabic benchmarks, alongside translation and summarization capabilities. The release gives researchers and developers an Arabic-focused foundation to adapt, with performance claims bounded by the report's comparisons.
Terms in this piece · Glossary
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
grounding — Tying a model's answers to checkable sources — retrieved documents, live data, tool results — instead of letting it answer from memory alone.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
articleArcee Meraj MaarjArcee AI
articleCan Large Language Models Review the Scientific Literature? A Test Against a 102-Article BenchmarkAbdullah Al Kalbani, David M. Schultz, Alexandra K. Anderson-Frey, Zhonghua Zheng, Catherine McGuire
articleA Systematic Evaluation of Cross-Lingual Consistency Enhancement Methods in Multilingual Language ModelsJirui Qi, Mingyang Wang, Hinrich Sch\"utze, Raquel Fern\'andez, Arianna Bisazza