Rakuten Group, Inc. has recently been selected to participate in the third term of the Generative AI Accelerator Challenge (GENIAC), an initiative backed by Japan’s Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). This collaboration is part of a larger effort to foster the development of generative AI within Japan.
The primary goal of the GENIAC project is to provide critical computing resources that facilitate the advancement of generative AI technology. By fostering collaboration around the latest trends and technologies, the program encourages knowledge sharing among developers, enhancing the landscape for AI innovation.
Since its inception, R&D support has been available for earlier terms of the program, with the first becoming accessible in February 2024 and the second in October 2024. Rakuten’s project was selected during the application phase of the third term, which commenced in March 2025. The company has been proactive in developing and releasing AI models optimized for the Japanese language within the open-source community since March 2024.
From the outset, Rakuten has emphasized cost efficiency in its AI initiatives, opting to create smaller, highly efficient models like Rakuten AI 2.0. This model employs a mixture of experts (MoE) architecture, allowing only relevant subsets to activate during query processing, resulting in significantly lower operational costs compared to traditional dense models.
As part of its commitment to the project, Rakuten is set to kick off research and development in August 2025 on an advanced open-weight AI foundation model. This model aims to incorporate groundbreaking techniques that will greatly enhance memory capabilities, thus improving information retrieval when generating responses. The end goal is to tackle existing limitations of generative AI models concerning memory recall and overall performance—a worthy challenge for any ambitious AI research team.
Looking ahead, Rakuten is paving the way for a more personalized AI experience, focusing on memory retention in user interactions. By enabling large language models (LLMs) to recall past conversations, Rakuten plans to foster long-term relationships with users empowered by predictive suggestions. This approach represents a significant leap beyond the constraints of current transformer architectures, which often grapple with maintaining extended context. Who knew memories could be a primary feature in AI?
In addition to bolstering memory, Rakuten plans to enhance operational efficiency through improved training and inference algorithms. These advancements will open new avenues for personalized AI applications across the expansive Rakuten Ecosystem, elevating customer experiences and streamlining business operations.
Yu Hirate, Vice General Manager at Rakuten Group’s AI Research Supervisory Department and Rakuten Institute of Technology Worldwide, expressed enthusiasm for this cutting-edge project, stating: “I am very pleased to be able to work on the development of a cutting-edge generative AI foundation model with the support of NEDO and the METI. Through this cost-effective AI model, we hope to contribute to the realization of AI agents that are best optimized for the Japanese language and are highly personalized, as well as empower local businesses and boost the economy.”
As it moves forward, Rakuten plans to leverage its vast data resources, extensive channels, and growth strategies to create new value not just in Japan, but for customers around the globe.
What is the Generative AI Accelerator Challenge (GENIAC)?
GENIAC is an initiative supported by METI and NEDO, focused on fostering the development of generative AI in Japan by providing essential resources and promoting collaboration among developers.
What innovative features is Rakuten introducing in its new AI model?
Rakuten’s advanced AI foundation model will incorporate enhanced memory capabilities, allowing it to better recall previous interactions, fostering a more personalized user experience.
How does Rakuten plan to improve AI efficiency?
Rakuten aims to enhance operational efficiency through advanced training and inference algorithms, which will unlock new opportunities for personalized AI throughout the Rakuten Ecosystem.
