rinna

What is rinna?

rinna is a Japanese AI company that builds language models tuned for Japanese language and culture. It publishes open models on Hugging Face and offers AI characters for businesses and creators.

Key capabilities of rinna

• Nekomata models: Continual pre-training of Qwen-7B on 30B Japanese and English tokens
• Youri models: Continual pre-training of Llama 2 7B on 40B Japanese and English tokens
• Instruction and chat variants: youri-7b-instruction and youri-7b-chat
• Open model releases: Published on Hugging Face under the rinna organization
• AI characters: Characters for businesses and creators
• Speech models: Japanese speech capabilities

How rinna works

rinna takes base models and applies continual pre-training on mixed Japanese and English text to improve Japanese performance. The resulting checkpoints are released on Hugging Face for developers to download and run, and business offerings build on that work. The company website could not be reached when reviewed, so product details come from its model pages.

Who uses rinna?

Developers and researchers working on Japanese NLP, and businesses seeking Japanese language AI or characters.

rinna pricing

The models are downloadable from Hugging Face, with license terms stated on each model card. Business and character services are not priced publicly.

rinna alternatives

Alternatives for document AI include Instabase and Hyperscience, and Stanford CoreNLP for classic NLP.