THaLLE
Finance-specialized Thai large language model from KBTG, the technology arm of Kasikornbank.
Overview
THaLLE (Text Hyperlocally Augmented Large Language Extension) is a Thai-language large language model developed by KBTG, the technology group of Thailand's Kasikornbank, specialized for the financial domain. It was developed and evaluated using the Investment Consultant exam dataset from the Stock Exchange of Thailand, scoring 72%, 72% and 84% on the P1, P2 and P3 exam levels. The work explores small, resource-efficient domain-specialized models and model merging, with weights released for research use.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 24 Jul 2026 and may be out of date or incomplete.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.
ASEAN Perspective
THaLLE in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
What this is for: A Thai-language LLM specialized in financial-domain understanding, benchmarked against Thai investment-consultant exams.
Who this is for: Thai fintech and financial-services developers and researchers building Thai-language finance NLP; it is a model, not personalized financial advice.
Availability: Open weights released by KBTG for research; self-hosted.
What people say
Published on arXiv by KBTG, THaLLE is a research artifact, not a commercial product you would find on G2 or Trustpilot. — the credible signal is its technical reports. KBTG published THaLLE on arXiv (with a follow-up "Thai Financial Domain Adaptation of THaLLE" report), and it appears on Papers With Code, which is the appropriate venue for a domain-specialized finance model of this kind.
The headline result is specific and checkable. On the Stock Exchange of Thailand's Investment Consultant licensing exam, THaLLE scored 72%, 72%, and 84% on the P1, P2, and P3 levels. That P3 score is notable for matching GPT-4-class performance on that specific test. The engineering story here is efficiency. The team used ReLoRA, continued pretraining, rsLoRA, and DPO to get a small model punching above its weight on Thai financial reasoning.
The claims come with significant caveats. The evidence is a single-institution technical report with no independent reproduction, and strong scores on one licensing exam don't establish broad or safe financial competence — this is a model, not licensed investment advice. Weights are released for research use rather than as a supported product, adoption beyond KBTG's own experiments isn't documented, and A benchmark win on a licensing exam is a poor proxy for reliability in finance, where hallucination risk is a deal-breaker.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including THaLLE's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with THaLLE 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 THaLLE directly →
Spotted something out of date? Suggest an update →
Alternatives to THaLLE
More in LLMs & Chat
