ByteDance trains massive AI model potentially rivalling Anthropic

Chinese technology giant ByteDance is training a new artificial intelligence model that could potentially rival the scale of advanced systems developed by U.S. companies like Anthropic. This move signals China's continued efforts to narrow the technological gap in the competitive AI landscape.

According to three individuals with knowledge of the project, ByteDance's model is in the early stages of training and could potentially include as many as 10 trillion parameters. If realized, this would be triple the size of Moonshot's Kimi K3, currently the largest Chinese-developed model publicly known.

The model is undergoing pre-training, a process typically lasting three to six months, before it is fine-tuned and potentially released. The final parameter count will be determined at a later stage of development, dependent on the success of the current phase.

This development underscores the global race to develop increasingly powerful AI models. Companies worldwide are investing heavily in large-scale AI training to unlock new capabilities and applications in the rapidly evolving field.