ByteDance trains massive AI model in bid to rival Anthropic

← Back to the feed

ByteDance trains massive AI model in bid to rival Anthropic

Ars Technica · 2 weeks ago

ByteDance is reportedly training an AI model that may reach up to 10 trillion parameters, positioning it to compete with leading US systems from Anthropic. The effort signals China’s push to close, and potentially surpass, the gap with American AI laboratories, although a model’s performance depends on training data and methods as well as parameter count.

The model is in pre-training, a process expected to take three to six months, and its final size has not been fixed. ByteDance’s Seed team has around 2,000 staff and is pursuing independently developed models rather than copying rivals’ outputs; the company has also expanded AI infrastructure, cloud services and chip ambitions. Its Doubao chatbot has 324 million monthly active users in China, while its SeeDance video model is considered globally competitive.

  • ByteDance may train a 10-trillion-parameter AI model.
  • China’s AI firms are narrowing the gap with US labs.
  • ByteDance is prioritising independent model development.

AI Technology Trending Weird & Viral

Read the full article at the source →