Syn Pro by Upstage and Karakuri Recognized as Domestic Foundation Model

On November 21, 2025, in Tokyo, Upstage AI Co., Ltd. (based in Minato, Tokyo, CEO: Hiroyuki Matsushita) announced a significant achievement: its Japanese-language-focused large-scale language model, Syn Pro, was officially recognized by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO) under the GENIAC project as a "Domestic Foundation Model". This recognition is awarded to models developed by companies and universities with development bases in Japan and is considered a foundational model developed autonomously within the country, encompassing full scratch development, additional learning, and fine-tuning.

Importance of Domestic Foundation Model for Japanese Entities

The recognition of Syn Pro as a domestic foundation model by METI represents a critical benchmark for Japanese companies and public institutions as they seek to adopt generative AI technologies. The growing emphasis on data sovereignty and compliance with domestic requirements has made the use of localized AI increasingly vital, especially in environments handling sensitive data.

Overview of Syn Pro

Syn Pro is a 31 billion parameter large-scale language model (LLM) co-developed by Upstage and Karakuri, which recently achieved a notable position by entering the top 20 of the Nejumi Leaderboard, standing as the only model outside of the United States and China as of November 20, 2025. By merging both organizations’ best practices, Syn Pro delivers superior accuracy in Japanese language processing even compared to larger models, expanding its application potential in sectors such as finance, insurance, and government that require high reliability.

Key Features of Syn Pro

1. Specialization in Japanese Language Accuracy and Context Understanding
Developed in collaboration with Karakuri, Syn Pro is optimized for Japanese data, domain-specific expressions, and cultural context. Its design considers Japan’s regulatory and operational requirements, allowing it to provide precise responses with minimal errors, even in specialized areas like law, finance, and insurance, while conforming to safety standards recognized by METI.

2. Complete On-Premise Operation Support
Unlike other models that often rely on external APIs, Syn Pro enables businesses to utilize AI on their internal networks or dedicated devices without exposing data to the outside. This flexibility allows for practical deployment, even in on-premises environments where immediate access through AI workstations and other local configurations is readily available.

3. Compact yet High-Performance
With a configuration of less than 32 billion parameters, Syn Pro has demonstrated Japanese language performance metrics that exceed larger frontier models in various benchmark assessments like the Nejumi Leaderboard. Its design negates the need for high-performance GPU clusters, enabling cost-effective implementation and operation within corporate environments.

Insights from Upstage's CEO

Hiroyuki Matsushita, CEO of Upstage, commented on the recognition, stating, "Syn Pro being added to the METI's list of domestic foundation models signifies a crucial advancement for our clients in Japanese enterprises and government agencies as they integrate AI tailored for their domestic operational needs. Partnering with Karakuri, we have thoroughly pursued an AI solution that is safe and reliable for use in Japan's business environments. This recognition will propel us to deliver high-performance AI infrastructure to a broader audience."

Future Developments

Syn Pro is poised to enhance its presence as an AI foundational tool across major Japanese industries such as finance, social infrastructure, public service, and manufacturing. Additionally, its role in addressing social issues and improving operational efficiency in government and safety domains of generative AI are among the anticipated future contributions of this domestic model.

About Upstage

Upstage is a global AI startup supported by Amazon and AMD, offering secure, high-performance, and cost-efficient generative AI technology. Its Japanese language-focused Syn series and Document AI solutions are being driven towards practical applications in sectors like finance, insurance, administration, and healthcare. Visit Upstage Website

Understanding Domestic Foundation Models

Domestic Foundation Models refer to those developed by companies or universities registered in Japan and possessing development facilities within the country. They include full scratch models, models enhanced through additional learning of existing frameworks, and models fine-tuned from domestic foundations.

For more details on the list and summaries of recognized domestic foundation models, visit GENIAC Prize.