Taiwanese tech giant Foxconn announced the launch of its first large language model, FoxBrain, on Monday, aiming to revolutionize manufacturing and supply chain management through advanced AI technology.
Developed using 120 Nvidia H100 GPUs within just four weeks, this marks a significant leap for the world’s largest contract electronics manufacturer.
FoxBrain is built on Meta’s Llama 3.1 architecture and represents Taiwan’s pioneering effort in creating a language model with reasoning abilities tailored for traditional Chinese and Taiwanese linguistic styles.
Despite a minor performance lag when compared to China’s DeepSeek distillation model, Foxconn claims that FoxBrain’s capabilities are nearly on par with international benchmarks.
Functionalities of the model span data analytics, decision support, document collaboration, mathematical reasoning, problem-solving, and code generation.
Meanwhile, Foxconn underlines plans to team up with tech partners to broaden the model’s applications, sharing its open-source knowledge and advancing AI utilization in manufacturing and supply chain management.
The tech giant stated that Nvidia lent its assistance via the “Taipei-1” supercomputer, situated in Taiwan, and provided technical consulting throughout the model’s training process.
Moreover, the company plans to unveil more information about the model at Nvidia’s GTC developer conference scheduled for mid-March.
