Rakuten Releases "Japan's Largest High-Performance AI Model," Its Configuration File Shows It's Based on DeepSeek V3 Architecture

Odaily News According to monitoring by 1M AI News, Rakuten Group has released the open-source model Rakuten AI 3.0, calling it "Japan's largest high-performance AI model." The model adopts a MoE architecture with a total of 671 billion parameters, activating 37 billion parameters per inference, a context window of 128K, optimized for Japanese scenarios, and outperforms GPT-4o in multiple Japanese benchmark tests.

This model is one of the achievements of the GENIAC project promoted by Japan's Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO), receiving partial computing power support. Rakuten did not disclose the source of the base model, only stating it was built based on achievements from the open-source community.

The community discovered in the HuggingFace model files that its config.json contains "deepseek_v3" and related architecture fields, and the parameter scale and context configuration are consistent with DeepSeek V3, indicating that this model may be based on a Japanese fine-tuning of DeepSeek V3.