Nice language fashions: Foxconn from Taiwan mentioned on Monday that he has launched his first massive language mannequin and intends to make use of expertise to enhance the manufacturing and provide chain administration.
The mannequin, referred to as “Foxbrain”, was skilled utilizing 120 of the H100 NVIDIA GPUs and accomplished in about 4 weeks, mentioned the biggest producer of contractual electronics in a press release.
The corporate, which assembles iPhone for Apple and in addition produces NVIDIA synthetic intelligence servers, mentioned the mannequin is predicated on LlamA three.1 Meta structure. It’s optimized for the primary Taiwan’s first massive language mannequin, with reasoning capabilities for conventional Chinese language and Taiwanese kinds, he mentioned.
Foxconn mentioned that though there was a slight efficiency hole in comparison with the Deepseek distillation mannequin in China, its basic efficiency could be very near world -class requirements. Initially designed for inner functions, Foxbrain covers knowledge evaluation, determination -making, doc collaboration, arithmetic, reasoning and downside fixing.
Foxconn mentioned he intends to collaborate with technological companions to increase the mannequin functions, to share their open-source info and to advertise you in manufacturing, to handle the provision chain and to make sensible selections. Nvidia supplied help by her “Taipei-1” supercomputer, primarily based in Taiwan, and supplied a technical consulting whereas making ready the mannequin, Foxconn mentioned.
Taipei-1, the biggest supercomputer in Taiwan, is owned and operated by Nvidia in Kaohsiung, a southern metropolis on the island. Foxconn will announce extra particulars in regards to the mannequin on the NVIDIA GTC builders convention in mid -March.