YTL AI Labs launches Malaysian LLM ILMUchat

YTL AI Labs has launched ILMUchat, a proprietary large language model designed around Malaysia’s multilingual and highly colloquial communication patterns. The model is optimised for rojak speech, where Bahasa Melayu, English, Chinese, local dialects and Manglish are often mixed within a single conversation.

The platform extends beyond text generation, with the pro version allowing users to build mini-applications and design presentation slides inside the ecosystem. YTL AI Labs trained the model on localised data and chat applications so it can understand references such as PTPTN and KWSP, as well as contextual distinctions like kuih lapis made in Kuala Lumpur or Sarawak. ILMUchat currently has over 10,000 organic users, with early usage focused on Malay-English translation, document understanding, summarisation and localised content generation.

YTL AI Labs is also positioning ILMUchat as infrastructure for data sovereignty, citing risks tied to foreign cloud-hosted APIs, the US Cloud Act and potential access restrictions during geopolitical or regulatory disputes. The company says its training strategy targets top-10 performance in open-source benchmarks for areas such as physics, biology and advanced coding, while each training run takes roughly 20 to 25 days across dedicated GPUs. Through Yes, ILMUchat Pro will be offered to over three million users and otherwise retails at RM50 per month.