A large-scale language model optimized for Japanese, developed by NII. Available in 8 sizes from 150M to 172B parameters. Trained on llm-jp-corpus v3. Supports SFT/DPO with Instruct2/Instruct3. Apache 2.0 license (for 13B and smaller).
Parameters
Undisclosed
Context Window
License
Proprietary
Release Date
2025-01-27
API Pricing
API pricing for this model is not yet available
Strengths
- •Comprehensive model family with multiple size options
- •Strong instruction-following and alignment capabilities
- •Open license for smaller models promotes accessibility
Weaknesses
- •Performance gap between smallest and largest models
- •Licensing restrictions on larger models may limit use
- •Alignment data may not cover all desired behaviors
Use Cases
- •Japanese conversational AI and customer support
- •Content creation and summarization in Japanese
- •Research on multilingual instruction tuning