A Japanese-optimized large language model developed by NII. Part of a family spanning 8 sizes from 150M to 172B parameters. Trained on llm-jp-corpus v3. Supports SFT/DPO via Instruct2/Instruct3. Licensed under Apache 2.0 (for 13B and smaller).
Parameters
Undisclosed
Context Window
License
Proprietary
Release Date
2025-05-27
API Pricing
API pricing for this model is not yet available
Strengths
- •Enhanced instruction following with 'instruct4' tuning
- •Compact size enables faster inference and lower cost
- •Open license facilitates deployment and modification
Weaknesses
- •The 1.8B scale limits capacity for complex reasoning
- •Iterative versions may have incremental, not revolutionary, gains
- •Lacks the depth of understanding of much larger models
Use Cases
- •Lightweight Japanese assistant for mobile applications
- •Quick response generation for Japanese FAQ systems
- •Rapid prototyping of Japanese conversational interfaces