Foxconn Launches FoxBrain LLM Based on Llama 3.1

Foxconn, the leading electronics contract manufacturer, has launched its first large language model (LLM), 'FoxBrain', to enhance manufacturing efficiency and supply chain management. Developed using 120 Nvidia H100 GPUs, FoxBrain was trained in four weeks and is based on Meta’s Llama 3.1 architecture, making it Taiwan's first LLM optimized for traditional Chinese and Taiwanese language styles. Although slightly behind China's DeepSeek in performance, FoxBrain is close to world-class standards and aims to be open-sourced to promote AI in manufacturing. This initiative underscores Foxconn's strategic shift towards AI, driven by the need for new revenue streams and the competitive AI landscape in China. Nvidia supported the development through its Taiwan-based supercomputer, 'Taipei-1', and further details about FoxBrain will be shared at Nvidia's upcoming GTC developer conference.