Rakuten Group has unveiled its latest AI innovation, Rakuten AI 3.0, a cutting-edge Japanese large language model (LLM) developed under the Generative AI Accelerator Challenge (GENIAC) project, supported by Japan’s METI and NEDO. This model boasts approximately 700 billion parameters and utilizes a Mixture of Experts (MoE) architecture, enhancing its efficiency by activating only a portion of its parameters for each task. Available via the Rakuten AI Gateway, this model promises significant cost reductions—up to 90% in tests—compared to competing AI models. Scheduled for a public release in Spring 2026, Rakuten AI 3.0 offers superior performance for Japanese language processing, refined through high-quality, proprietary data. With this launch, Rakuten aims to continuously advance its AI offerings and enrich user experiences, positioning itself as a leader in Japan’s AI landscape.