Thailand just launched its own large language model, positioning it as national infrastructure. It’s called ThaiLLM, and it signals a shift in how the country is thinking about AI: less as a tool to import, and more a technology to build and control domestically.

A model built for the Thai language

ThaiLLM was developed by the National Science and Technology Development Agency (NSTDA), working alongside the Ministry of Higher Education, Science, Research and Innovation, the National Electronics and Computer Technology Center (NECTEC), and other partners.

At its core, it’s a large language model trained specifically on Thai-language data, calling attention to the fact that most widely used AI systems today are built primarily around English or multilingual datasets where Thai is a smaller component. 

That often shows up in subtle ways that only a native speaker will find unnerving. As a highly contextual language, Thai can be difficult to nail. Cultural references don’t land, and tones can be all over the place.

By training on Thai-language data at scale, ThaiLLM aims to better understand how the language is actually used, including nuance, context, and cultural patterns.

The model was trained on more than 100 billion tokens and comes in two sizes, eight billion and 30 billion parameters. This gives developers flexibility depending on how they want to deploy it.

Keeping data closer to home

And one of the more strategic aspects of ThaiLLM isn’t the model itself, but where it runs.

It operates on Thailand’s supercomputer, ThaiSC, which allows data to be processed and stored domestically. That means government agencies, businesses, and individuals can use AI systems without sending sensitive data to servers outside the country.

This directly addresses a growing concern in AI adoption. Many organizations rely on foreign platforms, which raises questions around data control, compliance, and long-term dependency.

ThaiLLM is positioned as an alternative. Not necessarily replacing global models, but giving Thailand its own baseline infrastructure to build on.

AI tools beyond chatbots

ThaiLLM isn’t limited to chatbot-style interactions.

It’s designed to support agentic AI, meaning it can be used to build systems that operate with a degree of autonomy. This spans automation, decision-support tools, and workflows that go beyond simple prompts and responses.

In practical terms, this is where the model starts to matter for businesses. Rather than just answering questions, it can be integrated into internal systems, handling tasks, assisting decision-making, or powering customer-facing services in Thai.

Several Thai companies have already started working with ThaiLLM, including KBTG, SCB 10X, and research applications from VISTEC, particularly in experimental medical use cases.

In its initial phase, the model is available through four sub-models developed by both public and private organizations:

• OpenThaiGPT (AIEAT)
• Pathumma (NECTEC)
• Typhoon-S (SCB 10X)
• THaLLE (KBTG)

This kind of multi-entity rollout spreads development across different sectors, allowing the model to evolve through real-world use, rather than confining it to research.

ThaiLLM is accessible via API, which means developers can start building on top of it immediately. It’s also compatible with the OpenAI SDK, lowering the barrier for teams that already work with existing AI systems. Instead of starting from scratch, they can adapt current workflows to a locally hosted model.

There’s also a public-facing ThaiLLM Playground, where users can test the system through a chat interface. One notable feature is that responses include cited sources, aimed at improving reliability and transparency.

A different approach to AI development

What ThaiLLM does is it reflects a broader shift in how countries are approaching AI.

Rather than relying entirely on global platforms, there’s a growing push toward national AI systems that can handle local language, local data, and local use cases more effectively.

For Thailand, that means building something that fits how people actually communicate and how organizations operate, while keeping control over where data lives and how it’s used.

Whether ThaiLLM becomes widely adopted will depend on how well it performs in practice, but the direction is clear: AI is something Thailand is starting to build for itself.

source :