The National Science and Technology Development Agency (NSTDA) has partnered with Siam AI Corporation to launch a Thai large language model (LLM), marking a foundational step for Thailand's artificial intelligence (AI) capabilities and sovereign AI strategy.
The initiative is backed by a 300-million-baht investment from Siam AI.
Sukit Limpijumnong, president of NSTDA, said during the signing of a memorandum of understanding with Siam AI Corporation last Friday that the collaboration supports the development of the Thai LLM, a model which should understand Thai context and culture.
He said that having a native Thai LLM is a crucial component for the country's digital infrastructure, fostering the growth of AI development in the Thai language.
The project will play a vital role in advancing AI development tailored to Thailand's needs, fostering innovation in digital services, and enhancing the nation's competitiveness, said Mr Sukit.
NSTDA plans to support deep research and development of a sandbox for proof of concept, as well as cultivate a skilled workforce in collaboration with Siam AI to strengthen the research ecosystem in the long term, he said.
Ratanaphon Wongnapachant, chief executive of Siam AI, said the collaboration is a stepping stone for developing a Thai LLM to drive AI for commercial use in public and private organisations.
"We should develop Thai LLMs ourselves, as it is our native language, rather than relying on foreign-developed models. Otherwise, we risk losing valuable data that could be used for cognitive development, an essential foundation for the country," he said.
Foundation models or LLMs with multimodal capabilities are essential for advancing next-generation AI applications, such as autonomous vehicles, which require the integration of text, vision, audio and sensor data for real-time decision-making.
Siam AI hosts one of the most advanced high-performance computing infrastructures in Asia, featuring Nvidia Blackwell architecture and a liquid-cooled rack cluster designed to support LLMs, Mr Ratanaphon said.
Among Asian countries, Thailand is one of only four -- alongside China, Japan, and South Korea -- that have developed their own LLMs.
The Thai LLM is expected to feature 7 billion parameters and be completed within three months. The project is supported by a total investment of 300 million baht from Siam AI, building on an earlier development phase that had already received funding of 20–40 million baht.
Chai Wutiwiwatchai, executive director of the National Electronics and Computer Technology Center (Nectec), said the agency brings expertise in multimodal LLMs to support Siam AI, which has demonstrated strong capabilities in text-based LLM development.
Nectec developed Pathumma LLM, an AI model designed to support multimodal data processing -- including text, speech and images. It is specifically tailored to handle state and public sector documents, with a strong focus on understanding official language, structure, and contextual nuances.
Moreover, Nectec will deploy government AI projects on Siam AI's infrastructure, which supports scalability and flexibility for pilot initiatives.
Additionally, it will focus on developing a skilled workforce in high-performance computing to drive future advancements.
According to the Oxford Insights in Government AI Readiness Index 2024, Thailand ranked 35 out of 188 governments.
"Thailand is showing strong momentum in AI development, particularly in the medical sector, where we have deep expertise. This presents a promising opportunity for the country to export AI-driven services in the future," said Mr Chai.
In May, Siam AI signed an agreement with Naver Cloud, a South Korean company specialising in cloud computing and AI solutions, to develop a Thai LLM and AI agents for the tourism sector.
According to Naver Cloud, by leveraging their expertise in developing and operating LLMs and extensive Thai language datasets, the two companies aim to create a Thai-specific LLM by the end of this year.
This model will be designed for real-world applications, with plans to launch an AI agent tailored to Thailand's high-demand tourism industry.