ByteDance is reportedly developing a new large scale artificial intelligence model that could contain as many as 10 trillion parameters, highlighting the Chinese technology company’s continued investment in advanced AI research and infrastructure. According to a report published by the Financial Times, the project remains in the early stages of development but could eventually approach or even exceed the estimated size of Anthropic’s latest Mythos model. The report, which cited people familiar with the project, indicates that ByteDance is accelerating its long term AI ambitions as competition among global AI developers continues to intensify. ByteDance has not publicly commented on the report, while Reuters stated that it was unable to independently verify the information.
The report states that the new model is currently undergoing the pre training phase, which is typically one of the most resource intensive stages of AI development. Industry estimates referenced in the report suggest that Anthropic Mythos contains approximately 8 trillion parameters, while other prominent Chinese models remain considerably smaller. Moonshot AI Kimi K3 is estimated to have around 2.8 trillion parameters, while Meituan LongCat 2.0 and DeepSeek V4 Pro are each believed to contain approximately 1.6 trillion parameters. Although parameter count is often used to indicate the overall scale of an AI model, experts note that it does not directly determine performance. Factors including training methods, model architecture, data quality, optimization techniques, and fine tuning also play important roles in determining how effectively an AI system performs across different tasks. The pre training process for models of this size generally requires between three and six months before developers proceed to additional stages such as fine tuning and deployment preparation.
The reported project forms part of ByteDance broader investment in artificial intelligence technologies over the past several years. According to the Financial Times, the company’s Seed research team is responsible for multiple aspects of foundation model development, including pre training, post training, inference, memory systems, learning methods, and model interpretability. ByteDance also operates dedicated engineering teams responsible for distributed AI training and high performance inference infrastructure required for large foundation models. The report further states that the company has significantly expanded its investment in artificial intelligence by building additional data centers and recruiting AI researchers across China and international markets. ByteDance Seed team reportedly consists of approximately 2,000 members and is led by former Google DeepMind researcher Wu Yonghui. Alongside model development, the company has continued expanding its Volcano Engine cloud business, which provides enterprise AI services, while also pursuing plans to develop proprietary artificial intelligence chips that could support future computing requirements.
According to the report, ByteDance has adopted an independent approach to developing foundation models rather than relying on model distillation techniques commonly used within the industry. Model distillation generally involves training a smaller model to replicate the knowledge or outputs of a much larger model. The Financial Times reported that ByteDance founder Zhang Yiming believes building original AI capabilities is essential for achieving long term competitiveness instead of depending on technologies developed by rival AI laboratories. During an internal meeting held approximately two weeks before the report was published, Zhang reportedly encouraged the Seed team to remain focused on achieving world class artificial intelligence capabilities over the long term rather than concentrating on temporary competitive gaps. While the project remains under development and its final specifications have yet to be confirmed, the reported initiative reflects ByteDance continued efforts to strengthen its position in the rapidly evolving global artificial intelligence landscape through investments in research, infrastructure, enterprise AI services, and large scale model development.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
