ByteDance is reportedly in early talks to train a large language model with more than 5 trillion parameters — a scale that would put it far ahead of most known rivals.
If the plan goes through, the model would become China's largest known AI model, surpassing Alibaba's Qwen3.8-Max (2.4 trillion parameters) and Moonshot AI's Kimi K3 (2.8 trillion). But the project is still in its early discussion phase, and whether it actually gets trained or released is far from certain.
The effort is expected to be led by Xiang Liang, head of ByteDance's Seed Foundation, alongside Shen Ke, who leads LLM pretraining data. Seed is already reshuffling team responsibilities and allocating resources for the project…