TLDR: Rakuten Group has been selected for the third phase of Japan’s Generative AI Accelerator Challenge (GENIAC) project, spearheaded by the Ministry of Economy, Trade and Industry (METI) and NEDO. This initiative aims to bolster Japan’s generative AI development capabilities. Rakuten will commence research and development in August 2025 on a cutting-edge, open-weight AI foundation model, focusing on significantly expanding language model memory and enabling highly personalized AI agents.
Tokyo, Japan – Rakuten Group, Inc. announced today its selection for the third term of the Generative AI Accelerator Challenge (GENIAC) project, a significant national initiative promoted by Japan’s Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). The GENIAC project is designed to strengthen Japan’s overall generative AI development capabilities by providing essential computing resources and fostering knowledge exchange within the developer community.
The GENIAC project has been providing research and development support since its first term in February 2024, followed by a second term in October of the same year. Rakuten’s project was successfully selected through the third term application period, which commenced in March 2025.
Rakuten has been a proactive participant in the AI landscape, actively developing and releasing Japanese language-optimized AI models to the open-source community since March 2024. A core principle of their development strategy has been an emphasis on cost efficiency. This is exemplified by models such as Rakuten AI 2.0, which utilizes a Mixture of Experts (MoE) architecture. This innovative design allows the model to process queries by activating only relevant subsets of ‘experts,’ leading to a significant reduction in operational costs compared to traditional dense models.
As a newly selected participant in the GENIAC project, Rakuten is set to begin advanced research and development in August 2025. Their focus will be on a cutting-edge, open-weight AI foundation model. A key aspect of this research will involve integrating novel techniques designed to substantially expand the language model’s memory capabilities. This enhancement will effectively increase the volume of information the model can access when generating responses, thereby improving recall and overall performance.
Also Read:
- AWS to Launch AI Agent Marketplace, Featuring Anthropic and OpenAI Models
- EdgeRunner AI Partners with Air Force to Develop Secure Military AI Chatbot
Looking ahead, Rakuten is committed to pioneering the next phase of its AI journey, with a strong emphasis on personalization and memory for users. This involves developing Large Language Models (LLMs) that can ‘remember’ past interactions and conversations, moving beyond simple understanding to proactively suggest and cultivate long-term user relationships. This represents a significant advancement over the limitations of current transformer architectures, which often struggle with extended context windows. The ultimate aim of this research is to realize highly personalized AI agents that can offer more intuitive and effective user experiences.
