Foxconn has introduced “FoxBrain,” its first large language model designed for manufacturing and supply chain management.

Trained on 120 Nvidia H100 GPUs over four weeks, the model is based on Meta’s Llama 3.1 architecture.

It is optimized for traditional Chinese and Taiwanese language styles, with capabilities in data analysis, decision support, document collaboration, problem-solving, and code generation.

Initially developed for internal use, Foxconn plans to collaborate with partners to expand FoxBrain’s applications. Some open-source elements will be shared to enhance AI capabilities in manufacturing and supply chain operations.

The training was conducted using Nvidia’s Taipei-1 supercomputer in Kaohsiung.

More details about FoxBrain will be revealed at Nvidia’s GTC developer conference later this month.

Recent Foxconn developments