Rakuten AI 3.0: Japan’s Largest High-Performance AI Model Released

Tokyo, Japan – March 17, 2026 – Rakuten Group, Inc. Today announced the release of Rakuten AI 3.0, its latest large language model (LLM) and currently Japan’s largest high-performance AI model. Developed as part of the Generative AI Accelerator Challenge (GENIAC) project, a collaborative initiative between the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO), Rakuten AI 3.0 aims to accelerate AI development within Japan.

GENIAC Project and Development

The GENIAC project was established to bolster Japan’s capabilities in generative AI. Rakuten was selected for the third term of the project in July 2025, tasked with developing Japanese language-optimized AI models. Rakuten AI 3.0 builds upon a version initially unveiled in December 2025, undergoing further refinement and optimization. The project provides support for the substantial computing resources required for generative AI development, with part of the training costs for Rakuten AI 3.0 being covered by GENIAC.

Key Features and Capabilities

Rakuten AI 3.0 is optimized for the Japanese language and demonstrates strong performance in a variety of tasks, including writing, code generation, document analysis, and information extraction. Compared to Rakuten’s previous models, Rakuten AI 3.0 achieves significantly improved accuracy and robustness. The model is a Mixture of Experts (MoE) architecture with approximately 700 billion parameters, offering a superior grasp of Japanese language and culture. It also supports English.

Open Source Availability

Rakuten AI 3.0 is available free of charge under the Apache 2.0 license from the official Rakuten Group Hugging Face repository. This open-source approach is intended to foster collaboration and accelerate AI development across the Japanese technology landscape.

Rakuten’s Vision for AI Development

“Rakuten is committed to delivering high-quality, cost-efficient models that empower businesses and users,” stated Ting Cai, Chief AI & Data Officer of Rakuten Group. “Rakuten AI 3.0, our largest and most competitive model, is an outstanding combination of data, engineering and innovative architecture at scale. By sharing open models, we aim to accelerate AI development in Japan.”

Technical Specifications

• Model Architecture: Mixture of Experts (MoE)
• Total Parameters: 671 Billion
• Activated Parameters per Token: 37 Billion
• Context Length: 128K

Limitations and Bias

As with all large language models, Rakuten AI 3.0 has limitations and may generate biased, inaccurate, or unsafe outputs. Users are advised to exercise caution and implement appropriate safeguards when deploying the model in production environments.