Foxconn Reveals Its First Large Language Model

Foxconn, a leading Taiwanese company, announced the launch of its first large language model, "FoxBrain," aimed at improving manufacturing and supply chain management. The model, built on Nvidia's H100 GPUs over four weeks, is based on Meta’s Llama 3.1 architecture. It stands out as Taiwan's premier large language model equipped with reasoning capabilities tailored for traditional Chinese and Taiwanese language styles.

Although FoxBrain slightly lags behind China's DeepSeek's model, it nearly matches global standards. Initially intended for internal use, FoxBrain encompasses data analysis, decision support, document collaboration, mathematics, reasoning, problem-solving, and code generation. Foxconn plans to expand its applications through collaboration with tech partners, sharing open-source data, and advancing AI in manufacturing and supply chain management.

To support the development, Nvidia's supercomputer "Taipei-1" in Kaohsiung provided technical assistance during FoxBrain's training process. Further insights on the model will be unveiled at Nvidia’s GTC developer conference in mid-March.