rinna Co., Ltd. is a Japanese AI company best known for developing the AI character “Rinna.” Originating from the high-school-girl AI “Rinna” launched on LINE in 2015, the company was spun off from a Microsoft project in 2020. It has long focused on research and development of large language models (LLMs) specialized for Japanese, releasing many Japanese models as open source — including the 1.3-billion-parameter “japanese-gpt-1b,” GPT-2 and GPT-NeoX families, and the Qwen-based “nekomata” series. Today it offers the “Tamashiru” series of custom LLM solutions for enterprises and “Koemotion,” an AI voice service that generates emotional speech synthesis and face motion at the same time.
Key Features
- Japanese-specialized language model development: A pioneer that released GPT-family models trained on Japanese corpora early on, with continued R&D focused on Japanese conversational quality and naturalness
- Open-source model releases: Publishes many Japanese models for free on Hugging Face, including “japanese-gpt-1b” (1.3B parameters) and “nekomata” (Qwen-based 7B/14B models continually pre-trained on Japanese). Widely used by the research and developer community
- Expertise in AI character development: Builds on the casual-conversation and emotional-expression technology cultivated through “Rinna,” excelling at AI dialogue experiences with distinct character
- Koemotion (voice AI): Generates emotional speech synthesis and face motion simultaneously from text input alone. Available through both a browser UI and an API
- “Tamashiru” custom LLMs for enterprises: A solution providing LLMs customized to each company’s needs, designed to combine rinna’s own Japanese models with third-party LLMs
- Continuous research output: Regularly publishes research papers and models in multimodal AI spanning language, speech, and images
Pricing
| Offering | Price | Notes |
|---|---|---|
| Open-source models | Free | Published on Hugging Face; use under each model’s license terms |
| Koemotion | Free tier / paid plans | Free tier has a monthly request limit and no commercial use; paid plans combine a monthly base fee with usage-based billing |
| Tamashiru (enterprise) | Contact for quote | Individual estimates |
Pricing is as of August 2026. For the latest pricing, check the official site and rinna Developers.
Pros & Cons
✅ Pros
- A long track record in Japanese-specialized dialogue and text-generation models, with strong credibility in Japanese quality
- Many models are open source, so you can try them for free and run them in your own environment
- Provides AI experiences beyond text end to end — speech synthesis, face motion, and character dialogue
- As a Japanese company, support and business discussions are available in Japanese
⚠️ Cons
- Does not offer a general-purpose chat service for consumers like ChatGPT (mainly enterprise- and developer-oriented)
- Using the open-source models requires technical knowledge such as Python and GPU environments
- Enterprise solution pricing is not public, so inquiries are required before adoption
- Compared with the latest overseas frontier models, there is a gap in model scale and general capability
Comparison with Similar Services
| Comparison | rinna | ELYZA | CyberAgent (CALM) | Preferred Networks (PLaMo) |
|---|---|---|---|---|
| Provider | rinna Co., Ltd. | ELYZA, Inc. | CyberAgent | Preferred Networks |
| Distinctive strengths | AI characters, voice AI, Japanese LLMs | Llama-based Japanese LLMs, enterprise adoption support | Open-source Japanese LLMs | From-scratch Japanese LLMs |
| Open-source releases | Many | Yes | Yes | Yes (partial) |
| Voice AI | Yes (Koemotion) | No | No | No |
| Main offering formats | Enterprise solutions, API, OSS | Enterprise solutions | OSS, in-house services | Enterprise solutions, API |
Who Is It For
- Companies that want to integrate LLMs into their services with Japanese quality as the top priority
- Developers and researchers who want to try open-source Japanese models in their own environments
- People building AI experiences beyond text chat, such as AI characters and voice dialogue
- Developers who want easy API access to speech synthesis with emotional expression
Summary
rinna is a Japanese AI company that channels the accumulated Japanese-language AI research that began with the high-school-girl AI “Rinna” into two tracks: open-source model releases and enterprise solutions. It is not a chat service consumers can use right away, but it is a strong option for developers who want to run Japanese-specialized models themselves and for companies building AI experiences with character and voice expression. A good starting point is the public models on Hugging Face or Koemotion’s free tier.