Amazon Web Services (AWS) announced Monday that Malaysian startup Mesolitica has developed MaLLaM, a generative AI large language model (LLM) tailored for the Malaysian market.

Leveraging AWS’s cloud infrastructure, MaLLaM is capable of understanding Malay nuances, including slang, colloquialisms, and 16 regional languages, making it ideal for AI assistants across various industries.

The model is trained on 197 datasets and nearly 200 billion tokens of Malay-specific content and offers culturally relevant AI applications in customer service, content generation, and data analysis. With its advanced language processing capabilities, MaLLaM addresses diverse dialects from states such as Johor, Kedah, and Sarawak, enabling accurate, multilingual communication for businesses and government.

Moreover, by using AWS Trainium and AWS Inferentia chips, Mesolitica has reduced compute costs by 87% and increased training throughput by 5.5 times, ensuring MaLLaM delivers fast, cost-efficient AI performance.