離開
進入《TAIPEI TIMES》Hon Hai launches its first advanced LLM ‘FoxBrain’
The Hon Hai Precision Industry Co logo is pictured outside the company’s headquarters in New Taipei City’s Tucheng District on Feb. 17 last year. Photo: CNA
Staff writer, with CNA
Hon Hai Precision Industry Co （鴻海精密） yesterday said that its research institute has launched its first advanced artificial intelligence （AI） large language model （LLM） using traditional Chinese, with technology assistance from Nvidia Corp.
Hon Hai, also known as Foxconn Technology Group （富士康科技集團）, said the LLM, FoxBrain, is expected to improve its data analysis capabilities for smart manufacturing, and electric vehicle and smart city development.
An LLM is a type of AI trained on vast amounts of text data and uses deep learning techniques, particularly neural networks, to process and generate language. They are essential for building and improving AI-powered servers.
Nvidia provided assistance and consultation through its “Taipei-1” supercomputer, allowing the Hon Hai Research Institute to use the Nvidia NeMo AI service platform to complete the training of the FoxBrain LLM, Hon Hai said.
During the training of FoxBrain, the researchers used 120 Nvidia H100 graphics processing units （GPUs） and Nvidia Quantum-2 InfiniBand, the seventh-generation InfiniBand. The technology provided AI developers and researchers with the fastest performance available, Hon Hai said.
The training was completed in just four weeks at low cost and high efficiency, it said.
The FoxBrain model demonstrates strong comprehension and reasoning abilities, performs well in mathematics and logical reasoning tests, and is optimized to produce text that feels more natural and relevant to Taiwanese users, it said.
The model covers a wide range of areas, such as data analysis, decisionmaking aid, writing assistance, mathematics, code generation and problem-solving through its inferential capability, it added.
FoxBrain is expected to be made open-source, providing AI computing codes to other enterprises and researchers, enabling them to use LLM technologies to enhance and promote their own products, Hon Hai said.
FoxBrain has made significant progress in math tests compared with the Chinese-language model Taiwan Llama. It also outperforms Meta Platforms Inc’s current models of the same level in mathematical reasoning, the company said.
While FoxBrain is still a little behind DeepSeek’s （深度求索） models, its performance is already close to the best in the world, Hon Hai said.
FoxBrain is expected to drive the growth of smart devices and enhance AI-driven applications in manufacturing, supply chain management and intelligent decisionmaking, the company said.
Hon Hai is set to showcase FoxBrain at Nvidia’s GPU Technology Conference in San Jose, California, on Monday. The company would present its achievements in a keynote speech titled “From open source to frontier AI: Build, customize and extend foundation models,” it said.
新聞來源：TAIPEI TIMES
