ThaiLLM launched to strengthen Thailand's national AI capacity
Backed by NSTDA and partners, ThaiLLM offers 8B and 30B models, in-country processing on ThaiSC, API access and a path towards agentic AI.
- ThaiLLM is a large Thai-language AI model launched by Thailand's National Science and Technology Development Agency (NSTDA) and its partners to serve as national AI infrastructure.
- The model runs on a national supercomputer, ensuring that government, business, and personal data can be processed within the country, reducing dependence on foreign AI systems.
- Trained specifically on Thai-language data, it is designed to better understand local context and culture, and is accessible to developers via an API to encourage practical use in various sectors.
- Initial sub-models are already available from research agencies and private companies, with adoption underway in industries such as finance, technology, and medicine.
The National Science and Technology Development Agency (NSTDA) has launched “ThaiLLM”, a large Thai-language artificial intelligence model (Large Language Model: LLM) developed to serve as national AI infrastructure, through collaboration between the Ministry of Higher Education, Science, Research and Innovation (MHESI), NSTDA/ National Electronics and Computer Technology Centre (NECTEC) and DEF.
ThaiLLM has been developed for practical use with more than 100 billion training tokens.
It comes in two model sizes, 8 billion parameters and 30 billion parameters, and runs on the ThaiSC supercomputer, allowing all data to be stored and processed within the country.
More than 700 people have taken part in the project’s development, and it has been designed to support Agentic AI that can operate autonomously.
Its use is not limited to chat-based interaction, but can also be developed into automated systems and decision-support systems for organisations.
In its initial phase, ThaiLLM is available through four sub-models from Thai research agencies and private-sector companies: OpenThaiGPT-ThaiLLM-8B-Instruct-v7.2 by AIEAT, Pathumma-ThaiLLM-qwen3-8b-think-3.0.0 by NECTEC, Typhoon-S-ThaiLLM-8B-Instruct by SCB 10X, and THaLLE-0.2-ThaiLLM-8B-fa by KBTG.
A key strength of ThaiLLM lies in enabling government data, business data and personal data to be processed within the country, without having to be sent to foreign systems, reducing the limitations and risks associated with dependence on foreign AI.
At the same time, training the model specifically on Thai-language data helps the system understand local context, culture and patterns of language use more accurately.
ThaiLLM is also available via API, allowing developers to build on it immediately.
It supports a usage format compatible with the OpenAI SDK, making it easier to adapt to existing systems.
Thai private-sector companies have already started using ThaiLLM across several industries, including finance, technology and medicine.
These include KBTG, SCB 10X and its use in experimental medical work by VISTEC, reflecting the practical adoption of AI in the business sector.
It is also open for trial use through the ThaiLLM Playground platform, which supports chat-based use and displays answers with cited sources to improve the credibility of responses.
The launch of ThaiLLM reflects Thailand’s AI development direction, which aims to strengthen the country’s capabilities while keeping data under domestic control and opening access to practical-use technology for businesses and developers.
