JD.com has released xLLM, an open-source AI inference engine built using Chinese-made chips.

The Chinese ecommerce giant said xLLM uses a dynamic scheduler to prioritize requests, and a flexible architecture to optimize computing resources for AI tasks.

The engine is designed to handle complex, multimodal applications, and includes features like a multi-level pipeline execution engine and computational optimization tools.

JD.com said xLLM is already used in its retail business for applications such as AI assistants, customer service, and supply chain operations.

The company claims the system has improved efficiency by over five times, and cut machine costs by 90%, but these figures have not been independently verified.

🔗 Source: AI Base

Recent JD.com developments