Taiwan's First Financial LLM Launched; Banking Applications to Go Live by End of Year
Taiwan's first Financial Large Language Model (FinLLM) project has officially launched. Initiated by the Financial Technology Industry Alliance and led by CTBC, 16 banks are collaborating to build an AI model optimized for local regulations, aiming for a preliminary release in Q3 and a final launch by the end of the year.
📋 Article Processing Timeline
- 📰 Published: April 22, 2026 at 18:41
- 🔍 Collected: April 22, 2026 at 19:02 (21 min after Published)
- 🤖 AI Analyzed: April 23, 2026 at 14:55 (19h 53m after Collected)
Central News Agency
(CNA Reporter Su Ssu-yun, Taipei, 22nd) Taiwan's first Financial Large Language Model project has officially launched. It is expected to release a preliminary model based on banking industry knowledge by the end of the 3rd quarter and launch the final version by the end of this year. Financial Supervisory Commission (FSC) Chairman Peng Jin-lung stated today that it is expected to bring three major benefits: strengthening national sovereign AI applications, promoting technological upgrades in the financial industry, and advancing inclusive finance. This year's first phase focuses on banking models, and the next phase will expand to the insurance and securities sectors.
The Financial Technology Industry Alliance, established under the guidance of the FSC, held a press conference for the launch of the "Financial Large Language Model (FinLLM)" project today at FinTechSpace. Attendees included Peng Jin-lung, Minister of Digital Affairs Lin I-ching, Executive Secretary of the National Science and Technology Council's Office of Science and Technology Chang Chen-hao, Chairman of the Taiwan Financial Services Roundtable Tung Jui-pin, and several general managers and chairmen from the financial industry.
CTBC acts as the convener of this project. CTBC Chief Information Officer Chia Ching-kuang explained that this year is based on banking knowledge. The first version of the model will be available at the end of the 3rd quarter, and the final version in the 4th quarter. The cost for the first version will be between 40 million and 70 million NTD, shared jointly by the 16 banks in the alliance.
Peng Jin-lung pointed out that the financial services industry relies heavily on language and text understanding and processing. Especially in banking, capital markets, and the insurance industry, a high proportion of work has automation potential. This serves as the foundation for the immense demand to develop large language model application scenarios in the financial industry. However, finance is a highly regulated industry involving many local laws and regulatory requirements. It is not suitable to directly apply existing general-purpose LLMs. In other words, promoting a localized FinLLM is indeed necessary.
Peng expects that after the FinLLM goes online, it will bring three major benefits. First, it will strengthen national sovereign AI applications. By building Taiwan's FinLLM, local laws, regulatory practices, and local industry knowledge will be internalized into the model, reducing reliance on overseas models and enhancing data governance and cybersecurity capabilities. Second, it will upgrade the technology of the entire financial industry. This time, 16 banks are participating, combining investments from universities, the Institute for Information Industry, and tech vendors. This lowers the repetitive investment costs for the financial industry, helping to quickly build shared infrastructure and improve overall service quality and efficiency.
Third, it will promote inclusive finance and enhance public value. Future results can not only be applied to the operational scenarios of participating financial institutions but also spill over to other small and medium-sized financial institutions, and even general enterprises, educational institutions, and the public, strengthening the universality of financial knowledge.
Peng pointed out that this plan is also included in the new Top Ten AI Construction Plans. He thanked the Ministry of Digital Affairs for providing the sovereign AI corpus as a model data source, and noted that assistance from the National Science and Technology Council will be needed in many areas in the future. The plan is just beginning; by the end of this year, models based on banking knowledge can be launched, and the next phase will expand to insurance, securities, and other important financial sectors.
Minister of Digital Affairs Lin I-ching pointed out that global large language models are developing rapidly. However, if you ask international LLMs about Taiwan's financial industry, they might sound reasonable but actually cite foreign laws and hallucinate. By building Taiwan's FinLLM, it can answer regulations and questions according to the Taiwanese context. The model for this project is being trained by Asia Pacific Intelligence. He hopes the project can become a model and work hand in hand with the financial industry to step onto the international stage. (Editor: Pan Yi-ching)
(CNA Reporter Su Ssu-yun, Taipei, 22nd) Taiwan's first Financial Large Language Model project has officially launched. It is expected to release a preliminary model based on banking industry knowledge by the end of the 3rd quarter and launch the final version by the end of this year. Financial Supervisory Commission (FSC) Chairman Peng Jin-lung stated today that it is expected to bring three major benefits: strengthening national sovereign AI applications, promoting technological upgrades in the financial industry, and advancing inclusive finance. This year's first phase focuses on banking models, and the next phase will expand to the insurance and securities sectors.
The Financial Technology Industry Alliance, established under the guidance of the FSC, held a press conference for the launch of the "Financial Large Language Model (FinLLM)" project today at FinTechSpace. Attendees included Peng Jin-lung, Minister of Digital Affairs Lin I-ching, Executive Secretary of the National Science and Technology Council's Office of Science and Technology Chang Chen-hao, Chairman of the Taiwan Financial Services Roundtable Tung Jui-pin, and several general managers and chairmen from the financial industry.
CTBC acts as the convener of this project. CTBC Chief Information Officer Chia Ching-kuang explained that this year is based on banking knowledge. The first version of the model will be available at the end of the 3rd quarter, and the final version in the 4th quarter. The cost for the first version will be between 40 million and 70 million NTD, shared jointly by the 16 banks in the alliance.
Peng Jin-lung pointed out that the financial services industry relies heavily on language and text understanding and processing. Especially in banking, capital markets, and the insurance industry, a high proportion of work has automation potential. This serves as the foundation for the immense demand to develop large language model application scenarios in the financial industry. However, finance is a highly regulated industry involving many local laws and regulatory requirements. It is not suitable to directly apply existing general-purpose LLMs. In other words, promoting a localized FinLLM is indeed necessary.
Peng expects that after the FinLLM goes online, it will bring three major benefits. First, it will strengthen national sovereign AI applications. By building Taiwan's FinLLM, local laws, regulatory practices, and local industry knowledge will be internalized into the model, reducing reliance on overseas models and enhancing data governance and cybersecurity capabilities. Second, it will upgrade the technology of the entire financial industry. This time, 16 banks are participating, combining investments from universities, the Institute for Information Industry, and tech vendors. This lowers the repetitive investment costs for the financial industry, helping to quickly build shared infrastructure and improve overall service quality and efficiency.
Third, it will promote inclusive finance and enhance public value. Future results can not only be applied to the operational scenarios of participating financial institutions but also spill over to other small and medium-sized financial institutions, and even general enterprises, educational institutions, and the public, strengthening the universality of financial knowledge.
Peng pointed out that this plan is also included in the new Top Ten AI Construction Plans. He thanked the Ministry of Digital Affairs for providing the sovereign AI corpus as a model data source, and noted that assistance from the National Science and Technology Council will be needed in many areas in the future. The plan is just beginning; by the end of this year, models based on banking knowledge can be launched, and the next phase will expand to insurance, securities, and other important financial sectors.
Minister of Digital Affairs Lin I-ching pointed out that global large language models are developing rapidly. However, if you ask international LLMs about Taiwan's financial industry, they might sound reasonable but actually cite foreign laws and hallucinate. By building Taiwan's FinLLM, it can answer regulations and questions according to the Taiwanese context. The model for this project is being trained by Asia Pacific Intelligence. He hopes the project can become a model and work hand in hand with the financial industry to step onto the international stage. (Editor: Pan Yi-ching)