ThaiLLM
Thailand's national open-weight Thai LLM, led by the Big Data Institute
Overview
ThaiLLM is Thailand's national LLM project led by the Big Data Institute under MDES: ThaiLLM-8B and 30B base models continued from Qwen3 on ~63B tokens, with 8B instruct versions from AIEAT, SCB 10X, NECTEC and KBTG. Apache-2.0 weights free; hosted Playground and API free but non-commercial.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with ThaiLLM
- ThaiLLM-8B and ThaiLLM-30B base models
- Instruct versions: OpenThaiGPT, Typhoon, Pathumma, THaLLE
- Thai-English medical assistant and pre-screening models
- Free Playground and developer API
- Apache-2.0 open weights on Hugging Face
ASEAN Perspective
ThaiLLM in Southeast Asia
For organisations needing to keep data inside Thailand, ThaiLLM is the leading option. Launched in April 2026 by MHESI, NSTDA and NECTEC, it trains and runs on ThaiSC's LANTA supercomputer, meaning data is never processed abroad. For a local bank, hospital, or ministry, that data-residency guarantee can outweigh any performance benchmark from a foreign model. The model has been piloted by KBTG, SCB 10X and VISTEC. With open weights and a free API, it's also accessible to organisations that can't afford inference on a frontier model.
ThaiLLM gives Thailand a permissively licensed national model family, with Thai-tuned 8B and 30B bases and four instruction-tuned versions from respected Thai teams, plus medical fine-tunes developed with hospitals. It is built on Alibaba's Qwen3 rather than from scratch, and its own model cards report a smaller Thai training set than launch publicity claimed. The free hosted Playground and API come with terms allowing training on user inputs and banning commercial use of outputs, so serious users should self-host the Apache-2.0 weights. A valuable resource for Thai-language AI work in government, research and business.
What we found
ThaiLLM is Thailand's national large language model project, run by the Big Data Institute (BDI) under the Ministry of Digital Economy and Society with NECTEC, VISTEC and other partners, and funded by the Digital Economy and Society Development Fund. It was officially launched in April 2026. Its foundation models, ThaiLLM-8B and ThaiLLM-30B, are continued pre-trains of Alibaba's Qwen3 on about 63 billion tokens, including 31.5 billion of Thai web text and 8 billion of curated Thai data such as law, government documents and research articles, with a 32,768-token context; the 30B is a mixture-of-experts model. Government launch material spoke of more than 100 billion Thai tokens, but the published model cards give the 63 billion figure. On the ThaiExam benchmark the developers report ThaiLLM-8B scoring 0.483 against 0.314 for the Qwen3-8B base.
Four Thai organisations built instruction-tuned 8B versions on top: OpenThaiGPT by AIEAT, Typhoon by SCB 10X, Pathumma by NECTEC and THaLLE by KBTG. Later fine-tunes include a Thai-English medical assistant with tool use and a medical pre-screening model built on Google's MedGemma, developed with Siriraj hospital and public hospitals. Data contributors include the Administrative Court, the National Research Council, the Office of the Council of State, the National Library and Nation Group, and more than 700 students and researchers have been trained since late 2025.
The base and instruct weights are free under the Apache 2.0 licence on Hugging Face, though some later fine-tunes carry no licence tag. A free Playground and developer API need Google sign-in, but their terms let BDI use what users type to develop and train models and forbid commercial use of outputs without BDI's written consent, so organisations with privacy or commercial needs should self-host the open weights instead.
For Thai government agencies, researchers and companies it is a useful, permissively licensed starting point for Thai-language AI.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including ThaiLLM's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with ThaiLLM unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to ThaiLLM directly →
Spotted something out of date? Suggest an update →
Alternatives to ThaiLLM
More in LLMs & Chat
