There are several bahasa raya (LLM) models on the market today – including in English and Bahasa Malaysia. Adding to the offer, now a group of researchers in the UAE is present with an open source Arabic language offering, called Jais.
It is named Jais, in conjunction with the name of the highest mountain peak in the UAE. This project is a joint effort between several parties, including the university there that focuses on artificial intelligence.
This Jais language model comes with 13 billion parameters, which are a combination of Arabic and English. Combinations are used in training this model, due to the lack of data in Arabic at the moment. These 13 billion parameters were trained involving 116 billion Arabic-language data tokens and 279 billion English-language data tokens.
Jais' offer in the form of open source is expected to provide development in the arena of artificial intelligence in the Arabic language, at the same time bringing development that facilitates more than 400 million Arabic language users in the world.
For those of you who are interested, you can check the official Jais website to learn more about its offerings and technical papers, or check the Hugging Face website to see the Jais source code.