Foxconn Develops Taiwan’s First Large Language Model “FoxBrain” with AI Innovations for Smarter Manufacturing

Taiwanese electronics firm Foxconn has introduced its first large language model (LLM), "FoxBrain," intended at optimizing manufacturing and supply chain management.  The world's largest contract electronics manufacturer trained the model with 120 Nvidia H100 GPUs, completing the procedure in four weeks.  

FoxBrain is built on Meta's Llama 3.1 architecture and is Taiwan's first LLM with vast reasoning skills tailored to traditional Chinese and Taiwanese linguistic styles. While conceding a slight performance difference with China's DeepSeek distillation model, Foxconn underlined that FoxBrain's performance is approaching global standards.  

FoxBrain was originally designed for internal applications and now supports data analysis, decision-making, document collaboration, mathematics, reasoning, problem-solving, and code generation. Foxconn wants to engage with technological partners, offer open-source resources, and increase AI applications in manufacturing and intelligent decision-making.  

Nvidia became crucial in the project, providing technical support and access to its Taiwan-based supercomputer, Taipei-1—the largest in the country, located in Kaohsiung. Foxconn plans to release additional information on FoxBrain during Nvidia's GTC developer conference in mid-March.

This strategic initiative demonstrates Foxconn’s dedication to AI-driven innovation, reinforcing its position in the global technology landscape while boosting operational efficiency in the electronics industry.

This article is based on information from The Hindu