A Japanese-optimized large language model developed by NII. Available in eight sizes from 150M to 172B parameters. Trained on the llm-jp-corpus v3. Supports SFT/DPO via Instruct2/Instruct3. Licensed under Apache 2.0 for models up to 13B.
Parameters
Undisclosed
Context Window
License
Proprietary
Release Date
2025-03-12
API Pricing
API pricing for this model is not yet available
Strengths
- •Mixture-of-Experts architecture offers efficiency.
- •Potential for high performance with fewer parameters.
- •Designed for scalable Japanese language tasks.
Weaknesses
- •Complex architecture may increase inference cost.
- •Performance dependent on routing mechanisms.
- •Less common than standard dense models.
Use Cases
- •Efficient Japanese language model inference.
- •Research into novel model architectures.
- •Tasks requiring balanced performance and resources.