Malaysia’s Own AI Chatbot ILMU Speaks Rojak Malaysian Just Like We Do
Malaysia is about to introduce its first locally built large language model (LLM), called ILMU — short for Intelek Luhur Malaysia Untukmu, which means “Intellect Integrity Malaysia for You”. Developed entirely by Malaysians at YTL AI Labs, ILMU is not just made in Malaysia, but also trained to think and respond like a Malaysian.
Unlike global AI models that may miss local context, ILMU has been trained on Malaysian language and culture. It can understand the way Malaysians really speak — including “rojak Malaysian”, where English, Bahasa Melayu, and other local languages mix in the same sentence. The model can interact in three ways: through text, voice, and even images, making it a fully multimodal AI.
In tests using MalayMMLU, the first Bahasa Melayu multitask benchmark, ILMU reportedly outperformed well-known AI models such as GPT-4o, Deepseek-V3, and GPT-5. This benchmark is designed to measure how well AI understands and uses Malay in different types of tasks. ILMU also takes into account cultural norms, so its responses feel more relevant to Malaysians.
According to a report from SoyaCincau, ILMU is already available for businesses to integrate via API, allowing companies to build the AI into their own services. For the general public, ILMU Chat will open for early access on 16 September 2025, in conjunction with Malaysia Day. More details are available at the official website, ilmu.ai.
Source: [1] ILMU Chat, [2] SoyaCincau
Read more of our articles below!
