ByteDance Developing Massive AI Model to Rival Anthropic's Mythos

ByteDance, the parent company of TikTok, is training a new artificial intelligence model that could reach up to 10 trillion parameters. This scale would make it significantly larger than the Kimi K3 model from Moonshot and potentially exceed the size of Anthropic's most advanced system, Mythos 5, which industry estimates place at approximately 8 trillion parameters. The model is currently in the pre training phase, a process that typically lasts between three and six months before refinement and release.

ByteDance has invested heavily in AI infrastructure over the last three years, expanding its data centers and hiring a development team known as Seed. This team, led by former Google DeepMind scientist Wu Yonghui, consists of roughly 2,000 members. The company is also investing in its cloud computing unit, Volcano Engine, and intends to develop its own AI chips. Unlike some competitors, ByteDance has avoided model distillation, which is the process of compressing a large model into a smaller one. Founder Zhang Yiming believes that independent development is the only way to surpass competitors, although some observers suggest this approach has slowed the company's progress.

While ByteDance remains discreet about its closed models, its Doubao consumer model is currently the most popular in China with 324 million monthly active users. The pursuit of this mega model reflects a broader trend of Chinese laboratories attempting to close the gap with leading United States AI firms.