Japanese StableLM 3B

About

The Japanese StableLM 3B is a cutting-edge large language model developed by Stability AI, consisting of 3 billion parameters in a transformer decoder architecture akin to LLaMA. Designed specifically for Japanese language tasks, it excels in generating coherent text and following instructions while being culturally relevant. Trained on approximately 100 billion tokens from diverse datasets, including Wikipedia and OSCAR, it provides a rich linguistic foundation. Despite its capabilities, it may exhibit biases from its training data and requires fine-tuning for specific applications, especially in sensitive contexts.

Capabilities