Fujitsu Launches “Takane”, a Big Language Model for Businesses that Offers the Highest Japanese Language Proficiency in the World

Fujitsu today announced the release of Takane, a major language model (LLM) designed for enterprise use in secure private environments. Cohere Inc. Developed in collaboration with,⁽¹⁾ Takane represents a significant leap forward in generative AI capabilities by offering world-class Japanese language capabilities.

Fujitsu will integrate this new LLM into generative AI services at Fujitsu Kozuchi through Fujitsu Data Intelligence PaaS (DI PaaS), an all-in-one operations platform that is part of Fujitsu Uvance, Fujitsu’s solutions portfolio for cross-industry solutions. the social challenges it will present. Takane will be available worldwide starting September 30, 2024.

Takane achieved world-leading results on the Japanese General Language Comprehension Evaluation (JGLUE) assessment⁽²⁾Designed for corporate use in a secure and private environment. Fujitsu will offer Takane alongside its generative AI framework for enterprise, which includes tracking technology for generative AI that ensures outputs comply with laws, regulations and corporate rules, and knowledge graph extended RAG technology for referencing large-scale text. With this comprehensive approach, Fujitsu aims to create a Master that supports customers’ business transformation.

With the launch of Takane, Fujitsu is elevating its generative AI offerings at Fujitsu Kozuchi with a high-precision LLM specifically designed for secure private environments. This strategic move aligns with Fujitsu’s vision to seamlessly integrate generative AI into business operations.

Takane will also be available as part of an initiative to provide full support for customers’ productive AI journeys by combining the LLM with the Uvance Wayfinders consultancy service and wider Fujitsu Uvance offerings. With this comprehensive approach, Fujitsu aims to enable its customers to uncover new values ​​and address social problems by increasing productivity, creativity and innovation in their business activities.

Vivek Mahajan, Corporate Vice President, CTO, CPO, Fujitsu Limited, says:

“Takane represents a key part of Fujitsu’s AI strategy announced last year. Takane will empower companies in industries that demand the highest levels of security to harness the power of generative AI. Fujitsu is committed to creating new value for businesses by developing the latest artificial intelligence technologies. “This commitment fuels our three growth drivers: modernization, Fujitsu Uvance and consulting, all supported by our innovative technology.”

Corporate Vice President Yoshinami Takahashi, COO of Fujitsu Limited, says:

“Fujitsu Uvance supports our customers’ business growth and solving societal challenges through advanced decision-making powered by data and artificial intelligence. We are excited to bring Takane, a cutting-edge LLM specifically optimized for Japanese, to a wider audience through our DI PaaS platform. “We are committed to supporting customer business transformation by bringing the most advanced AI to market, not only through our own innovations, but also through collaboration with our global partners.”

Cohere Inc. Co-Founder and CEO Aidan Gomez says:

“We are excited to bring Takane’s advanced Japanese Master’s Programs to global companies. Our partnership with Fujitsu accelerates the adoption of AI in this critically important market by delivering secure, performant AI designed specifically for commercial use in Japanese and other languages.”

Increasing demand for specialized LLMs optimized for business needs in Japan

The adoption of generative AI is increasing globally across all industries, including marketing and customer service. However, general purpose LLMs are often offered through public cloud services and are therefore not suitable for use in industries that handle confidential data and in roles that require compliance with laws, regulations and industry rules. Additionally, features of the Japanese language, including mixed use of multiple character types, omitted subjects, and honorifics, also pose significant obstacles to general-purpose graduate studies, limiting their accuracy and reliability in business contexts. A highly accurate Japanese Master’s Diploma is particularly important for industries such as government, finance, healthcare and law, where even minor language errors can have serious consequences.

Takane’s Characteristics

1. Strong Japanese Language Abilities

Takane builds on Cohere’s LLM Command R+ and combines Fujitsu’s extensive knowledge of developing Japan-specific LLMs with Cohere’s expertise in creating task-specific language models. Takane will be offered exclusively by Fujitsu worldwide. Thanks to additional training and fine-tuning to improve Japanese language skills, Takane outperformed competitors with industry-leading performance on the JGLUE benchmark for understanding the Japanese language ⁽³⁾ in natural language inference (JNLI), reading comprehension (JSQuAD) tasks, and semantic comprehension and syntactic analysis, as evidenced by its top ranking in the Nejumi Master’s rankings3. Takane also includes Command R+’s multilingual support (covering 10 languages) and the ability to automate business processes.

2. Secure Private Environment

Takane is a medium-sized Master’s degree that can be used in a secure and private environment. This enables its safe use in industries where data breaches are a concern, such as the financial, manufacturing and security sectors, all of which work with sensitive data.

3. Mastering Through Fine-Tuning

Using company data, Takane can further specialize customer operations through fine-tuning and customization. Cohere’s RAG technology combines with Fujitsu’s extended RAG and generative AI inspection technologies to further facilitate compliance with laws and regulations, as well as industry and corporate codes. Fine-tuning and RAG technology make it possible to use Takane even in industries with specialized terminology and frequent revisions to legislation, such as the financial sector.

Fujitsu will offer Takane through its DI PaaS platform, thus creating a business application that combines data and artificial intelligence. DI PaaS gives organizations the power to combine vast, disparate data sources, both internal and external, and gather meaningful insights. DI PaaS enables seamless connectivity and data analysis across departments and industries, helping customers unlock new insights and solutions and driving data-driven innovation across organizations.

Fujitsu will unlock increased productivity and creativity and continue to support customers’ business transformation through AI/data integration.

The plan to propose Takane. Image Credit: Fujitsu

Takefumi Yamamoto, Chief Operating Officer, Group Deputy Chief Information Officer, Mizuho Financial Group, Inc., says:

“In fiscal 2023, we conducted joint trials with Fujitsu to explore the potential of generative AI in systems development and maintenance. Since fiscal 2024, we have been using Fujitsu Kozuchi to improve system quality by streamlining parts of the system development process. Now, with the launch of the LLM Takane initiative We are excited to unlock even greater possibilities. By leveraging the knowledge graph extension RAG and other advanced technologies, we will be able to use our internal knowledge much more efficiently. “We believe it will be a tool.”

Shoji Tanaka, Ph.D., Senior General Manager, Enterprise AI Strategy Division, Senior Executive Vice President, DX Innovation Center, Mitsubishi Electric Corporation, says:

“Takane seamlessly combines world-class technologies from Fujitsu and Cohere. “This powerful combination opens up a whole new realm of possibilities for the productive AI market in Japan, and I’m excited to see what the future holds for both companies.”

Yuki Shuto, Partner, Chief Strategy and Innovation Officer, Asia Pacific Technology, Media and Telecommunications Leader, Deloitte Tohmatsu Consulting LLC, said:

“Generative AI has moved beyond proof of concept across multiple organizations to full-scale integration across products, services, and operations. The key to success will be secure, reliable LLMs that allow organizations to share and leverage with confidence. Takane, advanced Japanese language “With its capabilities, it has tremendous potential to accelerate transformation within Japanese enterprises and government institutions. Deloitte Tohmatsu, in collaboration with Fujitsu, is committed to unlocking the full potential of productive artificial intelligence and contributing to the growth of Japan’s economy and society.”

RUTILEA, Inc. CEO Takafumi Yano says:

“Under our slogan ‘AI Made Easy’, we are committed to providing highly accurate AI services that will enable the integration of AI into all business processes. We are currently pursuing vertical AI initiatives in specific sectors, including central government, local governments, automotive industry, pharmaceutical industry and energy supply optimization. In 2024, we will launch an AI development platform business by establishing a GPU data center in Okuma Town, Futaba Prefecture, Fukushima Prefecture, Japan. ‘Takane’ has the potential to significantly accelerate the adoption of AI across various vertical industries in Japan. “We look forward to the day when we can contribute to its development and use.”

1. Cohere Inc: Headquarters: Toronto, Ontario, Canada; San Francisco, California, USA; Co-founder and CEO: Aidan Gomez
2. World-leading results were achieved on the Japanese General Language Comprehension Evaluation (JGLUE): Takane outperformed other LLMs in JGLUE, including large-scale, general-purpose LLMs delivered through cloud services. There was uncertainty in the accurate data (ground truth) for JNLI and JCoLA, so the accurate data were corrected by multiple annotators and measured as reference values ​​(as measured by Fujitsu and Cohere in September 2024).
3. Leaving competitors behind: On the Nejumi LLM Leaderboard 3, which evaluates LLM models’ Japanese language proficiency, Takane achieved the highest performance in the semantic comprehension category with a score of 0.862 and in the syntactic analysis category with a score of 0.773 (measured by Fujitsu and Cohere). in September 2024).
4. What are you doing? Japanese version of GLUE (General Language Comprehension Assessment) (JGLUE: Japanese General Language Comprehension Assessment; Kentaro Kurihara (Waseda University), Daisuke Kawahara (Waseda University), Tomohide Shibata, 28th Annual Meeting of the Japan Language Processing Society (NLP2022))
5. JSTS: Task to estimate semantic similarity between a pair of sentences.
6. JCoLA: Task for syntactic evaluation of Japanese language models.
7. What? Natural language inference task that recognizes the inference relationship between a premise sentence and a hypothesis sentence.
8. JCommonsenseQA: Task to assess common sense judgment ability.
9. : Machine reading comprehension task to assess the ability to read a document and answer questions about it.