Rakuten has been selected for the third phase of the Generative AI Accelerator Challenge (GENIAC). This initiative is supported by Japan’s Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). The initiative’s goal is to foster the development of generative AI systems in Japan by providing computing resources and technical exchange.
The GENIAC initiative started in February 2024 and is divided into several phases. The third phase, starting in March 2025, emphasizes enhancing domestic research capabilities for large language models (LLMs). This phase is crucial for advancing infrastructure and algorithmic innovations. Rakuten plans to initiate its research and development by August 2025.
The company’s primary focus will be developing a Japanese language foundation model with open weights. This model will incorporate techniques to expand memory capacity, improving data retention and retrieval over long interactions. By addressing existing constraints in generative models, Rakuten hopes to enhance AI performance with improved contextual memory.
Previously, Rakuten released Japanese-language models to the open-source community, including their AI 2.0 model. This model employs a Mixture of Experts architecture, activating only relevant parts during inference and lowering computational demands compared to dense models. In this new development phase, the Japanese company intends to test model enhancements for extended interactions and consistent outputs.
In the future, Rakuten aims to create LLMs capable of retaining user-specific context, facilitating personalized AI interactions. This plan includes studying training and inference techniques to support customized AI use across various Rakuten platforms. The focus is on automation and improved customer interaction support.
Yu Hirate, vice general manager at Rakuten Group’s AI research department, expressed enthusiasm for working on the development of a cutting-edge generative AI model. “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 Ministry of Economy, Trade and Industry. 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,” Hirate said.
