Rakuten builds memory-enhanced Japanese AI model with government backing

Rakuten has been selected for the third phase of Japan’s government-backed Generative AI Accelerator Challenge (GENIAC), a program supported by the Ministry of Economy, Trade and Industry and the New Energy and Industrial Technology Development Organization (NEDO). The Japanese tech giant plans to develop an open-weight Japanese language model with enhanced memory capabilities, positioning itself to create more personalized AI applications across its business ecosystem while contributing to Japan’s domestic AI research capacity.

What you should know: Rakuten will focus on creating a lightweight, memory-augmented Japanese language model using a Mixture of Experts architecture starting in August 2025.

• The model aims to address existing technical constraints in generative AI systems that operate with limited contextual memory.
• Rakuten previously released Japanese-language models to the open-source community, including Rakuten AI 2.0, which uses MoE architecture to reduce computational overhead compared to standard dense models.
• The new development phase will test enhancements enabling models to handle extended interactions and deliver more consistent outputs across multiple user prompts.

In plain English: Think of current AI models like someone with short-term memory who can only remember the last few sentences of a conversation. Rakuten wants to build an AI that can remember much more of what you’ve discussed, making it better at personalized conversations over time.

The big picture: GENIAC’s third phase, which began in March 2025, focuses on improving domestic research capacity for large language models through infrastructure and algorithmic advancements.

• The initiative launched in February 2024 and provides computing resources and technical exchange opportunities to support generative AI development in Japan.
• Rakuten’s participation represents a strategic move to strengthen Japan’s position in the global AI race while developing culturally and linguistically optimized solutions.

Why this matters: The project could enable more personalized AI interactions by allowing models to retain user-specific context over longer periods.

• Rakuten intends to deploy these models across its various platforms, focusing on automation and customer interaction support.
• The development aligns with Japan’s broader strategy to reduce dependence on foreign AI systems while building domestic capabilities.

What they’re saying: Yu Hirate, vice general manager at Rakuten Group’s AI research supervisory department, emphasized the significance of government support for cutting-edge AI development.

• “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,” Hirate stated.
• “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.”

Broader applications: Rakuten has already begun commercializing AI solutions for business clients through its existing platforms.

• In January, Rakuten Mobile launched “Rakuten AI for Business,” a generative AI solution designed specifically for Japanese business culture, regulations, and industry norms.
• The service provides document generation, translation, idea development, analysis, and research capabilities through an interactive chat interface while operating in a secure environment that protects sensitive data.