Rakuten Group has announced the launch of its first Japanese large language model (LLM) and small language model (SLM), named Rakuten AI2.0 and Rakuten AI2.0mini.
The release of these two models aims to advance artificial intelligence (AI) development in Japan. Rakuten AI2.0 is based on a mixture of experts (MoE) architecture and is an 8x7B model, consisting of eight models, each with 7 billion parameters, acting as individual experts. When processing an input token, the system sends it to the two most relevant experts, selected by a router. These experts and the router are continuously jointly trained on a large volume of high-quality Japanese-English bilingual data.
Rakuten AI2.0mini is a new dense model with 1.5 billion parameters, designed for cost-effective deployment on edge devices suitable for specific application scenarios. It is also trained on mixed Japanese-English data, aiming to provide convenient solutions. Both models have undergone instruction fine-tuning and preference optimization, with a base model and an instruction model released to support businesses and professionals in developing AI applications.
All models are licensed under the Apache 2.0 license, and users can access them in the official Rakuten Group Hugging Face repository. Commercial uses include text generation, content summarization, question answering, text understanding, and dialogue system construction. Additionally, these models can serve as a foundation for further development and applications.
Cai Ting, Chief AI and Data Officer of Rakuten Group, stated: “I am incredibly proud of how our team has combined data, engineering, and science to launch Rakuten AI2.0. Our new AI models offer powerful and cost-effective solutions that help businesses make smart decisions, accelerate value realization, and unlock new possibilities. By open-sourcing these models, we hope to accelerate AI development in Japan and encourage all Japanese companies to build, experiment, and grow, fostering a collaborative and win-win community.”
Official Blog: https://global.rakuten.com/corp/news/press/2025/0212_02.html
Key Points:
🌟 Rakuten Group launches its first Japanese large language model (LLM) and small language model (SLM), named Rakuten AI2.0 and Rakuten AI2.0mini.
📊 Rakuten AI2.0 is based on a mixture of experts architecture, featuring eight expert models with 7 billion parameters each, dedicated to efficiently processing Japanese-English bilingual data.
🛠️ All models are available in the Rakuten Hugging Face official repository for various text generation tasks and can serve as a foundation for other models.
