Project & Service

Project Introduction

On September 16, 2023, the Shenzhen Research Institute of Big Data, in collaboration with The Chinese University of Hong Kong (Shenzhen) (CUHK-Shenzhen) and King Abdullah University of Science and Technology (KAUST) in Saudi Arabia, launched AceGPT—an open-source large language model specifically designed for the characteristics of the Arabic language, featuring leading performance. By the end of 2024, AceGPT will offer models in multiple sizes, including 7B, 13B, 32B, and 70B. In terms of Arabic, Chinese, and English capabilities, this model significantly outperforms competitors—specifically the Jais model developed in the UAE—making it the world's leading open-source Arabic large model. Additionally, AceGPT's Arabic capabilities surpass those of GPT-3.5 (175B) and are close to GPT-4 (1.7T?).

Research Focus

1. Model Training: This includes the collection and cleaning of localized data, the construction of localized instructions, and localized RLAIF.

2. Model Function Development: This encompasses multimodal features, long text processing, vocabulary expansion, and function call capabilities.

Main Outputs

1. Three top-tier conference papers.

2. Applications for national-level projects in both China and Saudi Arabia.

Specific Application Scenarios and Functions

The project serves Chinese enterprises expanding into the Middle East by providing them with Arabic large model technology, facilitating the localization of domestic software and hardware products in the Middle Eastern market.

Collaboration Model

1. API Provision  

2. Project Collaboration  

3. Establishment of Joint Ventures  

chat.acegpt.org