Rakuten Group has been selected for phase three of the Generative AI Accelerator Challenge (GENIAC), a Japanese government-supported initiative aimed at bolstering domestic research and development in generative artificial intelligence. The program, backed by Japan´s Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO), seeks to accelerate large language model research through the provision of advanced computing resources and structured technical exchange. Rakuten´s inclusion follows an application process for GENIAC´s third phase, which kicked off in March 2025 with a focus on fostering algorithmic and infrastructural innovations for Japanese language models.
Starting August 2025, Rakuten will begin developing a new Japanese-language foundation model with open weights, leveraging a Mixture of Experts architecture to deliver lightweight, memory-augmented capabilities. The project is designed to enhance the memory capacity of generative models, overcoming current restrictions related to contextual retention and sustained interaction. Rakuten intends to test improvements that will allow the new models to handle extended and multi-turn user prompts, improve consistency, and facilitate personalized outputs tailored for various business applications. This aligns with the company´s longer-term vision of building large language models that can retain user-specific context for highly individualized experiences across its digital platforms.
Rakuten has previously open-sourced models such as Rakuten AI 2.0, which also utilizes a Mixture of Experts structure to optimize computational efficiency. As part of its broader artificial intelligence strategy, the company earlier launched ´Rakuten AI for Business´, a generative artificial intelligence solution for enterprise clients focusing on business process automation, document generation, translation, and data security—features specifically tuned to Japanese regulatory and cultural standards. Notably, all user inputs to the platform are processed in a secure environment, with strict controls on data use and privacy, ensuring corporate customers can restrict sensitive information from being utilized without permission.
Yu Hirate, vice general manager overseeing Rakuten´s artificial intelligence research division, expressed optimism about the initiative´s contribution to creating highly personalized, Japanese-optimized artificial intelligence agents and its potential to empower local businesses and support national economic growth. The GENIAC project marks a key step for Japan in building an indigenous generative artificial intelligence ecosystem, with Rakuten at the forefront of technical research and practical deployment for language-specific models in commercial settings.
