ByteDance Trains 10 Trillion-Parameter AI Model to Rival Anthropic

ByteDance is training a new artificial intelligence model featuring up to 10 trillion parameters. The project is designed to surpass the scale of Anthropic's Mythos system, which industry estimates place at approximately 8 trillion parameters.

The model is currently in the pre-training phase, a process expected to last between three and six months before the company begins fine-tuning. This initiative follows a directive from founder Zhang Yiming, who urged employees to avoid relying on AI distillation techniques for short-term gains.

The development is part of a larger trend among Chinese technology firms to accelerate model release cycles. By increasing scale, ByteDance aims to match the capabilities of U.S. competitors in the large language model race, following similar efforts by other Chinese labs such as Moonshot AI, which recently released the 2.8 trillion-parameter Kimi K3 model.