Lightblue leverages Qwen to develop Japanese-focused LLMs and assistant
Lightblue Co., Ltd. used Alibaba Cloud’s Qwen as the base foundation model for Japanese language development. The company built and released its Karasu and Qarasu LLMs and planned to improve its Lightblue Assistant RAG service using Alibaba Cloud infrastructure.
- Organization
- Lightblue Co., Ltd.
- Industry
- Tech & Comms
- Location
- Japan
- Published
- December 2023
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Lightblue Co., Ltd.
- Provider
- Alibaba
- Maturity
- Unknown
- Linked source
- Alibaba Cloud Customers
No explicit deployment-stage evidence found.
Primary read
Use case focus
Showing 3 of 3
- 1AI model training
- 2Conversational assistants
- 3RAG infrastructure
- The company selected the open-source edition of Tongyi Qianwen (Qwen) after comparing other foundation models.
- It fine-tuned Qwen for Japanese understanding and used Alibaba Cloud Elastic Compute Service, Server Load Balancer, and Object Storage Service to deploy and scale the application.
- Qwen helped Lightblue release its Karasu and Qarasu models around December 2023.
- The company reported improved accuracy for Japanese language processing after fine-tuning.
Architecture
Lightblue fine-tuned Alibaba Cloud’s open-source Qwen foundation model for Japanese language understanding and deployed the application using ECS, SLB, and OSS. The company also positioned its Lightblue Assistant RAG service for future improvement and reduced reliance on third-party APIs.
Sources & evidence1
- Customer explicitly identified
- Primary source available
- Technical implementation details available
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?
Community
Comments
No published comments yet.