Foxconn, widely recognized for assembling Apple’s iPhones, has made a significant leap into artificial intelligence with the launch of its new language model, ‘FoxBrain.’ This model, based on Meta’s Llama 3.1 architecture, is specifically designed for traditional Chinese and Taiwanese languages.
It represents a notable achievement as Taiwan’s first AI language model. Developed quickly in just four weeks using 120 cutting-edge Nvidia H100 GPUs, FoxBrain is touted for its strong performance, particularly in data analysis, problem-solving, code generation, and decision-making tasks. While Foxconn acknowledged that FoxBrain slightly lags behind China’s DeepSeek model, it claims that the new model approaches global standards and is equipped for complex reasoning tasks.
Initially, Foxconn aims to use FoxBrain to enhance its internal operations, focusing on improving manufacturing efficiency, optimizing supply chain management, and facilitating smarter decision-making processes. Additionally, Foxconn plans to work with tech partners and make its AI research available through open-source platforms to promote wider industry collaboration and innovation.
Nvidia has played a crucial role in the development of FoxBrain, providing the necessary infrastructure and expertise through Taiwan’s powerful supercomputer, ‘Taipei-1,’ located in Kaohsiung. Foxconn plans to share more information about FoxBrain at Nvidia’s GTC developer conference later this month, signaling its ambition to lead in both hardware manufacturing and AI technologies.
